基于多层CatBoost的电力系统暂态稳定评估

Transient Stability Assessment of Power System Based on Multi-layer CatBoost

随着大规模新能源并网以及新装置的不断应用,电力系统暂态稳定问题日益复杂,为进一步提升暂态稳定评估(transient stability assessment,TSA)的精确性和可靠性,提出一种基于多层CatBoost的TSA方法。首先,以电力系统故障前的稳态运行变量作为输入特征,采用一种最大相关最小冗余(maximal relevance minial redundancy,mRMR)集成方案,从输入特征中筛选出多组不同的关键特征集。然后,利用这些关键特征分别单独训练多个CatBoost模型,建立多个CatBoost驱动的TSA模型,并结合多个训练好的CatBoost模型构建TSA综合模型。在暂态稳定分析时,综合多个CatBoost模型的分析,通过多数投票表决方式判定最终评估结果。最后,在IEEE 39节点系统和某省级电力系统上进行性能测试实验。测试结果表明:所提出的TSA综合模型不仅具有极高的预测精度,而且拥有良好的泛化能力和鲁棒性。

With the integration of large-scale new energy into the grid and the continuous application of new devices,the transient stability problem of power system is becoming increasingly complex.In order to further improve the accuracy and reliability of transient stability assessment(TSA),a TSA method based on multi-layer CatBoost was proposed.Firstly,the steady-state operation variables before the power system faults were taken as the input features,a maximal relevance and minimal redundancy(mRMR)ensemble scheme was adopted to select multiple sets of different key feature sets from the input features.Then,these key features were used to train multiple CatBoost models separately to establish the TSA model driven by multiple CatBoost,and combined with multiple trained CatBoost models to construct TSA comprehensive model.In the transient stability analysis,the final evaluation results were determined by majority voting based on the analysis of multiple CatBoost models.Finally,the performance test experiment was carried out on the IEEE 39-bus system and the certain provincial power system.The test results show that the proposed TSA comprehensive model not only has high prediction accuracy,but also has good generalization ability and robustness.