The Prediction Of Coal Gas Under One Mine Based On The Improved Lyapunov Exponent

Coal provides powerful energy support for the development of our national economy. But the production of coal is always plagued by the security problems. And in our country, gas accident is one of the major disasters that thread the safety production of one mine. So it has great significance to forecast the mine gas in a scientific and accurate way. A variety of factors affects the coal mine system. And the interaction of various factors forms a complex coal gas dynamical system which has chaotic properties. So it’s feasible to establish prediction model of underground coal mine gas concentration based on the chaos theory in order to achieve the purpose of early warning, prevention and control of coal gas accidents.According to the symbol selection of the prediction model of the Largest Lyapunov exponent based on the chaos theory, introduce a weighted local thought to derive a new prediction formula and verification and analysis the prediction model with single gas sensor data. Next, according to the multi variables selection problem of the prediction of multi sensor data, introduce the correlation analysis to find the strength between the variables. And then, improve the prediction model of the largest Lyapunov exponent by considering both cosine and Euclidean distance to obtain the nearest point of the prediction center point. At last, establish the prediction model of the improved largest Lyapunov exponent with the multi sensor data, then analysis and verification through the experiment.The main research contents of this paper are as follows:First, explore the technology of phase space reconstruction, which is associated with the chaos theory. And then using the correlation method and the irrelevant method to analyze and obtain the reconstruction parameters separately. Then study the identification of chaos, especially through whether the largest Lyapunov exponent is greater than zero to judge the system is chaos or not.Second, improve the prediction model, which is based on the largest Lyapunov exponent, of the single sensor data. And solve the problem of symbol selection through the using of weighted local method and the reduction of formula. And then use the real-time series data of the coal mine of Lu Tai Shan to forecast the gas value and analysis the results. The contrast experimental results show that the symbol of the improved forecasting model is certain with the root mean square error 2.61%, which is lower than the traditional 4.27%. It shows that the improved model is better in the gas density forecasting.Third, introducing the method of correlation analysis, to find which are the key factors of the gas concentration from lots of environmental factors firstly. Then use the close factors as the inputs of the multivariable prediction model. Then improve the prediction model, which is considering both cosine and Euclidean distance to obtain the nearest point of the prediction center point. And then use the multivariate real-time sensor data of the coal mine of Huoerxinhe to forecast the gas value and then make the contrast experiment with the prediction model of traditional methods and BP neural network methods. The results show that the improved model is more accurate than the contrast forecasting models. It illustrates that the improved model is more effective in the prediction of multi variable gas concentration.