面向变电站防汛的熵权分配混杂风险评估

Hybrid Risk Assessment of Entropy Weight Allocation for Substation Flood Control

针对传统电网防汛评估方法主观性强、适应性差和准确性低的问题,提出一种基于熵权分配的变电站实时防汛风险混杂评估模型。首先,根据多维静、动态数据对变电站防汛风险评估的影响程度,分别引入轻量级梯度提升机与长短期记忆神经网络构建变电站防汛静、动态数据评估子模型;然后,利用熵权法分配权重实现两子模型的有效组合;最后,进行对比实验,实验结果显示本文模型优于主流的单一评估模型和组合评估模型。算例分析表明,所提变电站防汛风险评估算法具有较强的适应性与较高的准确性。

Aimed at the problems of strong subjectivity,poor adaptability and low accuracy of the traditional power grid flood control assessment methods,a real-time flood control hybrid risk assessment model for substations based on entro⁃py weight allocation is proposed.First,according to the difference in the influences of multi-dimensional static and dy⁃namic data on the substation flood control risk assessment,the LightGBM and LSTM algorithms are introduced to con⁃struct sub-models for substation flood control static and dynamic risk assessment.Then,the entropy weight method is used to allocate weights to achieve an effective combination of the two sub-models.A contrast experiment shows that the proposed model outperforms the mainstream single and hybrid assessment models.The result of an example shows that the proposed substation flood control risk assessment algorithm has a strong adaptability and a high accuracy.