| 引用本文: | 邬春明,任继红.基于人工智能的暂态稳定裕度精细化预测[J].电力自动化设备,2021,41(12): |
| WU Chunming,REN Jihong.Refined prediction of transient stability margin based on artificial intelligence[J].Electric Power Automation Equipment,2021,41(12): |
|
| 摘要: |
| 人工智能算法在暂态稳定评估中得到了很好的应用。然而，电力系统是时变大系统，训练数据无法涵盖所有工况，模型需要在有限时间内更新；电力系统中稳定样本数远大于失稳样本数，导致模型对失稳样本学习不足。针对以上2点，提出了基于人工智能的暂态稳定裕度精细化预测方法。该方法将改进的极限梯度提升(XGBoost)树与双XGBoost回归树集成，平衡了2类样本数量差异对模型的影响，并实现了裕度预测。当运行工况变化较大时，结合增量学习技术，以较少的样本和较短的时间对模型进行有效更新。在2套IEEE系统上的实验结果表明所提方法可应用于暂态稳定评估。 |
| 关键词: XGBoost算法 暂态稳定评估 增量学习 电力系统 代价敏感 |
| DOI：10.16081/j.epae.202108008 |
| 分类号:TM712 |
| 基金项目:国家自然科学基金资助项目(61901102) |
|
| Refined prediction of transient stability margin based on artificial intelligence |
|
WU Chunming1,2, REN Jihong2
|
|
1.Key Laboratory of Modern Power System Simulation and Control & Renewable Energy Technology, Ministry of Education, Northeast Electric Power University, Jilin 132012, China;2.Department of Electrical Engineering, Northeast Electric Power University, Jilin 132012, China
|
| Abstract: |
| The artificial intelligence algorithm has been well applied in transient stability assessment. However, the power system is a time-varying system, so the training data cannot cover all the working conditions, and the model needs to be updated within a limited time. The number of stable samples in power system is much larger than the number of unstable samples, which leads to the lack of learning from unstable samples. In view of the above two points, the refined prediction method of transient stability margin based on artificial intelligence is proposed. This method integrates the improved XGBoost(eXtreme Gradient Boos-ting) tree with the double XGBoost regression tree, so that the effects on the model caused by the difference number of two types of samples are balanced, and the margin prediction is realized. When the operating conditions change greatly, the incremental learning technology is combined to effectively update the model with fewer samples and shorter time. Experimental results on the two IEEE systems show that the proposed method can be applied in transient stability assessment. |
| Key words: XGBoost algorithm transient stability assessment incremental learning electric power systems cost sensitive |
