论文部分内容阅读
径流预测历来是水利部门的一项重要工作,针对水库和河流中长期径流预测精度不高,提出了自适应调节人工蚁群算法(ARACS),对RBF神经网络参数进行优化,建立了自适应调节人工蚁群-RBF神经网络组合算法(ARACS-RBF)预测模型,综合考虑影响径流预变化因素,对安康水库进行中长期径流预测。对预测效果进行检验,结果证实该模型可真实地反映河川径流变化的总体趋势,并为判断时间序列数据的非线性提供了一种新方法。与RBF神经网络模型、人工蚁群-RBF神经网络模型预测结果进行对比,结果表明,应用ARACS-RBF模型对中长期径流量进行预测,预测精度更高、效果更好。该方法克服了RBF神经网络和人工蚁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RBF神经网络的泛化能力,收敛速度快,输出稳定性好,提高了径流预测的精度,置信度为98%时的预测相对误差小于6.5%。可有效用于水库和河川中长期径流预测。
Runoff prediction has always been an important work of the water conservancy department. In view of the low precision of medium and long-term runoff prediction in reservoirs and rivers, ARACS is proposed to optimize the parameters of RBF neural network, and adaptive adjustment Artificial ant colonyRBF neural network combined algorithm (ARACS-RBF) forecasting model, considering the factor of influencing runoff pre-change, forecasting the long-term runoff of Ankang Reservoir. The test results show that the model can truly reflect the overall trend of river runoff changes and provide a new method for judging the nonlinearity of time series data. Compared with RBF neural network model and artificial ant colony-RBF neural network model, the results show that ARACS-RBF model is applied to forecast medium and long-term runoff, and the prediction accuracy is higher and the effect is better. This method overcomes the shortcomings that the RBF neural network and the artificial ant colony algorithm are easily trapped in the local extreme, the search quality is poor and the precision is not high, the RBF neural network improves the generalization ability, the convergence speed is fast, the output stability is good and the runoff is improved The accuracy of prediction is less than 6.5% when the confidence is 98%. It can be effectively used for medium and long-term runoff prediction in reservoirs and rivers.