引水工程边坡位移 SABO-LSTM监测模型及风险状态研究

文章利用减法平均优化算法（subtraction-average-based optimizer，SABO）对长短期记忆网络（long short-term memory，LSTM）模型进行优化，建立SABO-LSTM引水工程边坡位移监测模型，实现了对边坡位移规律的准确预测。并以边坡各测点位移和位移速率为评估指标，根据各指标数据特征采用置信区间法对其进行风险状态分级，基于区间数距离计算得到各指标的基本概率分配；然后结合D-S（Dempster-Shafer）证据理论对各指标基本概率进行融合，利用融合结果对边坡位移风险状态进行评估，并结合监测模型预测结果，实现了对边坡位移风险状态的预评估。该研究结果可为边坡位移预测及风险状态评估提供技术参考。

In this paper, the subtraction-average-based optimizer (SABO) is used to optimize the long short-term memory (LSTM) model, and the SABO-LSTM-based slope displacement monitoring model of water-division project is established, thus achieving the accurate prediction of slope displacement law. The displacement and displacement rate of each measuring point of slope are taken as evaluation indexes. According to the data characteristics of each index, the confidence interval method is used to classify the risk state, and the basic probability assignment (BPA) of each index is calculated according to the interval number distance. Then, the basic probability of each index is fused with the Dempster-Shafer (D-S) evidence theory, and the risk state of slope displacement is evaluated by using the fusion results. Combined with the prediction results of the monitoring model, the pre-evaluation of the risk state of slope displacement is achieved. The research results can provide technical reference for slope displacement prediction and risk state assessment.