论文部分内容阅读
为了提高水库和河流中长期径流预测精度,提出了弹性自适应人工鱼群算法(RAAFSA)。应用RAAFSA算法训练BP神经网络,实现BP神经网络参数优化,形成弹性自适应人工鱼群-BP神经网络混合算法(RAAFSA-BP),对石泉水库进行中长期径流预测。仿真计算表明,弹性自适应人工鱼群优化的BP神经网络算法收敛速度快于BP神经网络算法、人工鱼群-BP神经网络算法和RBF神经网络算法。该混合算法克服了BP神经网络和人工鱼群算法易陷于局部极值、搜索质量差和精度不高的缺点,改善了BP神经网络的泛化能力,输出稳定性好,预报精度显著提高,每次预测相对误差绝对值都小于6%,合格率达到100%。该算法成功地解决了石泉水库中长期径流预测精度不高的难题,可有效用于水库和河川中长期径流预测。
In order to improve the accuracy of long-term runoff prediction in reservoirs and rivers, a flexible adaptive artificial fish swarm algorithm (RAAFSA) is proposed. The RAAFSA algorithm is used to train the BP neural network to optimize the parameters of the BP neural network, and a flexible adaptive artificial fish-BP neural network hybrid algorithm (RAAFSA-BP) is formed to predict the mid-long term runoff of the Shiquan Reservoir. The simulation results show that the BP neural network algorithm based on elastic adaptive artificial fish swarm optimization has faster convergence rate than BP neural network algorithm, Artificial Fish Stock-BP neural network algorithm and RBF neural network algorithm. The hybrid algorithm overcomes the shortcomings that the BP neural network and the artificial fish swarm algorithm are easily trapped in local extremum, the search quality is poor and the precision is not high, the generalization ability of the BP neural network is improved, the output stability is good, and the forecasting accuracy is significantly improved. The relative error of sub-prediction is less than 6% and the pass rate is 100%. The algorithm successfully solves the problem that the accuracy of long-term runoff prediction in Shiquan reservoir is not high enough, and can be effectively used for medium and long-term runoff prediction of reservoirs and rivers.