The Research Of Gas Emission Prediction Based On MPSO-LS-SVM Algorithm

Coal mine gas emission is one of the most dangerous disasters in the disasters of mine gas, the prevention and control of the disasters of mine have become the priority among priorities, as the safety work in mine. Due to gas emission factors are dynamic non-linear, between these factors and gas emission exist uncertain and non-linear. It brought certain difficulties in prediction of gas emission. Therefore, how to build has stronger generalization ability of the model to predict gas emission accurately, this is the problem need to be solved, which still exist in the research on prediction of gas emission in coal mine.In this paper, aiming at the need of safety in production, with the influence of mine gas emission factors as the object of study, with the gas emission prediction as the purpose, to study the particle swarm algorithm and least squares support vector machine theory and methods. Proposed a combination algorithm which used the modified particle swarm optimization algorithm and least squares support vector machine to creat the predictive model of gas emission. Using least squares support vector machine instead of support vector machine to predict gas emission, which reduces the computational complexity and improves training efficiency. Least squares support vector machine parameters to select differently, for the model training time and the prediction error will produce different effects, proposed the use of particle swarm optimization algorithm to optimize its width penalty factor C and nuclear factor σ. And then analyzes the operating mechanism of standard particle swarm optimization algorithm and the various parameters on the effect of algorithm performance, for particle swarm optimization algorithm easy to fall into local optimum drawback, the paper to improve the inertia weight about particle swarm optimization algorithm, and introducing the factor of convergence, proposed the modified particle swarm optimization algorithm. The algorithm overcomes the local convergence while accelerates the convergence speed, more able to adapt to complex optimization problems. Using the modified particle swarm optimization of least squares support vector machines to predict the gas emission, and use the monitoring to the gas emission of historical data in simulation, the experimental results show that this model and the standard particle swarm optimization of least squares support vector machines model, compared with BP neural network model, the convergence speed is fast, the convergence precision is high, the prediction error is low. Shows that the method has good feasibility and effectiveness, it is able to predict gas emission accurately.