改进BP神经网络的煤与瓦斯突出预测

Dangerous Forecast of Coal and Gas Bursting Based on Improved Back Propagation Neural Network

研究煤与瓦斯突出预测问题,煤与瓦斯突出是一种非常复杂的地质灾害,受到瓦斯、地应力、煤物理力学性质等多种因素影响,存在复杂的非线性系统问题,引起预测准确性差。传统BP神经网络存在收敛速度慢、易陷入局极小等缺陷。为了有效提高煤与瓦斯突出的预测精度,提出一种改进BP神经网络的煤与瓦斯突出预测算法。首先采用附加动量调整网络的权值,加快网络收敛速度,然后自适应调整网络学习速度,减少迭代次数,最后对煤矿的煤与瓦斯突出进行仿真,相对于传统BP神经网络,不仅有效地减少了迭代次数,加快了学习速度,而且提高了预测精度,为煤矿灾害的准确预测提供了依据。

Outburst of coal and gas is a kind of very complex geological disaster. In order to improve the predic- tion precision of coal and gas outburst, we proposed a prediction algorithm of coal and gas outburst based on the im- proved BP neural network. First, additional momentum was used to adjust the weights of the network ans accelerate the network convergence speed. Then the network learning speed was adaptively adjusted to reduce the number of it- erations. Finally, the coal and gas outburst simulation experiment was performed. Compared with the traditional BP neural network, the improved one effectively reduces the number of iterations, accelerates the learning speed, and improves the prediction precision.