基于粒子群优化神经网络的概率积分法预计参数的确定

为有效确定概率积分法预计参数,提高预计值的精度。将粒子群优化(PSO)算法和BP神经网络进行融合,采用改进的混合粒子群优化算法优化神经网络的权值和阈值。在分析概率积分法参数与地质采矿条件之间关系的基础上,建立了基于PSO优化BP神经网络的概率积分法预计参数的优化选择模型。以我国典型的地表移动观测站资料为例,将计算结果与实际值进行了对比分析,并与文献1]中改进BP算法进行了比较。结果表明,PSO-BP神经网络方法用于概率积分法预计参数的选取是可行的,收敛速度更快,计算精度更高。

Optimal selection of prediction parameters for probability-integral method using particle swarm optimization and BP neural network

In order to determine the prediction parameters of probability-integral method and to improve the prediction accuracy, a new method by combining particle swarm optimization algorithm (PSO) and BP neural network (PSO-BP) is presented. In this method, an improved hybrid PSO algorithm is used to optimize the connection weights and thresholds values of BP neural network. An optimal selection model for prediction parameters of probability-integral method using this hybrid PSO-BP neural network algorithm is constructed based on the relation analysis between the parameters of probability-integral method and geological mining conditions. Typical data of surface moving observation stations is used as samples. Comparison analysis is addressed between calculated values generated by PSO-BP method and observed values, and the performance of PSO-BP method is also compared with improved BP algorithm mentioned in first bibliography. Results indicate that PSO-BP calculation model has higher convergent speed and higher precision. A new optimal selection model for parameters of probability-integral method is provided.