Short-term Prediction of Linz-Donawitz Gas Generation Tendency Based on SVD-NCF-GA-BP

摘要

The prediction of Linz-Donawitz Gas (LDG) production and consumption tendency was paramount important in gas balancing and scheduling since it's an important secondary energy which each process in the steel and iron enterprise needed. Therefore, this paper proposed a prediction method combining curve fitting and GA optimized BP neural network to predict LDG short-term production trend. Specifically, proposed method firstly utilized SVD decomposition to preprocess instantaneous values of LDG production in order to extract a standard type of LDG production during a smelting cycle. Then the standard type was curve fitted to attain function formulas of the overall recovery about time series and meanwhile a series of function clusters and values were procured. Afterwards, GA optimized BP neural network was employed to train parameters of function clusters and thus a recovery trend of LDG during a production period was obtained, which was also called the prediction of short-time production trend. Finally, the actual data from a certain steel and iron enterprise was adopted to verify feasibility and efficiency of the proposed method, the results showed that proposed method had a good performance in predicting short-term LDG generation trend.