基于改进BP神经网络的海堤渗压监测模型研究

Multipoint seepage pressure monitoring model for sea dikes based on improved BP neural network

针对传统BP算法存在的收敛速度慢、易陷入局部极小等缺陷,采用附加动量算法和修正激活函数方法对网络模型进行改进。以BP神经网络为建模工具,将前期潮位因子、降雨因子和时效因子作为输入,多个渗压测点值为输出,建立了海堤渗压多测点监测预报模型。将原始网络模型、改进网络模型的预测值与实际测值进行对比分析表明,改进的BP神经网络模型在海堤渗压监测中具有更快的收敛速度和更优的预测精度。

In view of the existing defects of traditional BP neural network such as slow convergence speed, easy to fall into lo- cal minimum and so on, additional momentum algorithm and modified activation function method are used to improve the original neural network model. By taking BP neural network as a modeling tool and choosing the antecedent tidal -level factor, rainfall factor and time factor as the input, and the multiple osmotic pressure value as the output, we build a muhipoint seepage pressure monitoring model for sea dikes. The predicted values obtained respectively by the original model and improved model and meas- ured values are compared. The results show that the improved BP neural network has faster convergence speed and better accura- cy.