基于粒子群算法优化BP神经网络的CO_2通量预测

A CO_2 flux prediction model based on particle swarm BP neural network algorithm

针对现有准确地预测CO_2通量方法的不足,提出了一种以粒子群算法和反向传播(BP)神经网络相结合的预测模型。为了防止粒子种群的快速趋同效应,引入了自适应变异算子。通过对陷入局部最优的粒子进行变异操作,提高了算法的寻优性能。利用粒子群算法得到BP神经网络的初始权值和阈值,对优化后的BP神经网络和普通的BP神经网络分别创建CO_2通量预测模型。实验结果表明,基于粒子群改进BP神经网络模型能较好表达CO_2通量与主要因素之间的非线性关系,相对于一般BP神经网络具有更好的非线性拟合能力和更高的预测准确性。

Aiming at the shortcomings of the existing prediction methods, a prediction model combining particle swarm algorithm and back propagation （BP） neural network is proposed. An adaptive mutation operator is introduced in to avoid link effect. Mutation operation on particle swarm which is falling into lo- cal optimum will improve the optimization performance. The initial weight value and threshold of BP neural network will be obtained using the improved particle swarm algorithm. CO2 flux prediction models are established for optimized and traditional BP neural network respectively. Experimental results show that the improved BP neural network model based on particle swarm optimization （PSO） can better express the nonlinear relationship between CO2 flux and the main factors, which has better nonlinear fitting ability and higher prediction accuracy than traditional BP neural network.