基于改进BP神经网络模型的地面沉降预测及分析

Prediction and Analysis of Land Subsidence Based on Improved BP Neural Network Model

针对区域性地面沉降问题,用遗传算法优化BP神经网络的初始权重,建立了地面沉降预测模型.该模型克服了BP神经网络模型存在的收敛速度慢、易陷入局部极小点的缺点.采用后验差检验法对模型拟合结果进行了检验,结果表明模型具有很好地拟合与泛化能力.应用该模型对地下水位影响强度进行了分析,表明地面沉降与地下水位存在一致响应趋势.

In order to control land subsidence efficiently, a coupling model of genetic algorithm and back-propagation （BP） neural network was applied to the simulation of land subsidence, aiming at overcoming shortcomings of the BP neural network model, such as falling into local minimum value easily and being slow in convergence. The coupling model passed the posterior-variance-test and good fitting and generalization were obtained. The results calculated through the proposed model indicate that the variation of land subsidence rate in the researched district has consistent tendency with underground water level.