基于BP神经网络技术的区域短期地震预测模型研究

地震预测是一个世界性科学难题,特别是短期与临震预测的水平与社会需求相距甚远。论文在详细分析研究地震数据特征以及常规地震预测方法的基础上,提出了一种可以实现地震震级量化预测的新方法,此方法通过解算出地震参数和天文时变参数并建立地震预测模型,对未来预测周期内发生的最大地震震级进行量化预测。本文以实验区域为研究对象并选取6个月为预测周期,采用线性回归分析方法和常规BP神经网络方法进行研究。经回溯检验,其地震震级预测中误差分别为±0.78级和±0.61级,精度均有待提高。经过总结上述两种方法的优缺点,创新的提出了基于线性回归与神经网络技术的地震预测融合模型,回溯检验结果表明,融合模型的震级预测中误差为±0.41级,地震预测效果显著提高。

Regional Short-term Earthquake Prediction Model Based on BP Neural Network

Earthquake prediction is a worldwide scientific problem,especially the prediction level for short-term and imminent earthquake.Based on detailed analysis and induction of the seismic data and their characteristics,a method which gives the quantitative prediction for earthquake magnitude is introduced in this paper.By this method,after calculating the earthquake parameters and the astronomical time-varying parameters,an earthquake prediction model can be established to give the quantitative prediction for earthquake magnitude in the future prediction period.In this research,the research object is the experimental areas,the prediction period is 6months,and Linear Regression analysis and conventional BP(Back Propagation) Neural Network have been used alone in prediction.Through backtracking test,the RMSEs(root mean square error) of earthquake magnitude prediction are ±0.78 Ms and ±0.61 Ms.Then after summarizing the advantages and disadvantages of the two methods,a integration model,based on linear regression and neural network,has been proposed.Through backtracking test,the RMSE of earthquake magnitude prediction reached ± 0.41 Ms,which is result that it improving significantly.