Research On The Coupling Mine Hydraulic Discharge Prediction Model Of Time Series And BP Neural Network

The mine hydraulic discharge prediction is a question must be faced and answeredfor the development and production of the mine, and it’s a basic work for thedevelopment and safety in production. Artificial neural network and time series modelare two important mine hydraulic discharge prediction models, but there are still somelimitations in practical applications. In order to improve the applicability and predictionaccuracy, an ideal time series function which included long-term trend and periodic itemwas constructed, and then the function values were simulated and predicted by themethods of BP neural network and time series model respectively and the relative errorswere5.82%and3.40%. After analyzing the advantages and disadvantages of the twomodels, the method through coupling the time series and BP neural network to improvethe applicability and prediction accuracy was put forward. After analyzing thecombination method of the coupling model, a serial coupling model of time series andBP neural network was built. The long term trend, periodic term, seasonal variation andirregular changes were decomposed and extracted with multiplication decompositionmodel of time series and the values of long term trend, seasonal variation and irregularchanges were got by the methods of trend extrapolation, season average method and BPneural network respectively. The periodic term was fit and forecasted with the Fourierseries approximation method when it was obvious, otherwise with the BP neuralnetwork method. The main programs of the coupling model based on MATLAB weredesigned according to the homologous prediction methods, the functions and theparameter selection principle. After calculating, the forecast relative error of thecoupling model was2.13%and improved the accuracy of the prediction compared withthe single model above. Finally, the coupling model was used to predict the waterinflow of longmen colliery. The prediction results showed that, the forecast relativeerror of the coupling model was1.4818%and the forecast relative error of BP neuralnetwork was5.2176%. The coupling model had a great degree of improvement inforecast accuracy. The research displayed that the prediction coupling model of BPneural network and time series had better applicability and effectiveness, and can beused as a new forecast method of the mine hydraulic discharge.