基于AI技术的电网关键稳定特征智能选择方法

An Intelligent Key Feature Selection Method of Power Grid Based on Artificial Intelligence Technology

随着全国联网规模的扩大,电网时空特性日趋复杂,全面把握电网运行特性显得至关重要,仅靠人工经验手动筛选特征难以做到快速和准确。因此依托人工智能(artificial intelligence,AI)技术快速高效地找到能够反映电网安全信息的关键特征,对于实际电网的实时监控和安全运行具有重要意义。针对上述要求,该文提出一种电网关键稳定特征智能选择方法。该方法借助信息论和数据挖掘技术,采用分段方式集成改进(mutual information feature selection,MIFS)方法和封装方法进行电网关键稳定特征的智能选择。第一阶段利用改进MIFS方法进行特征初筛;第二阶段采用封装方法和后向搜索策略进一步选择关键特征。该分段智能选择方法针对电力系统安全稳定评估应用场景实现多种方法的集成应用,一方面可以有效地减小特征维度,实现特征的智能筛选,而且将运行经验纳入考虑范围,符合调度人员的先验认知,便于调度运行人员实时监控;另一方面减小了所选特征集合的冗余度,同时实现了特征分区筛选,提高了计算效率,有利于实时动态安全评估的开展。在IEEE-39节点系统上的仿真算例验证所提方法的有效性。

With the expansion of the nationwide power grid network, the complexity of space and time of power grid is increasing. It is difficult to fully grasp the operating characteristics of the power grid based on human experience only. Therefore, finding the key features that can reflect the security information of the power grid quickly and efficiently based on artificial intelligence(AI) technology is of great significance to the monitoring and safe operation of the power grid. Aimed at the requirements above, this paper presented an intelligent feature selection method for power grid. This method uses information theory and data mining technology to intelligently select key features with integration of improved mutual information feature selection(MIFS) and wrapper method. On the first stage, improved MIFS method was used to preliminarily select features;on the second stage, wrapper method and backward search strategy were adapted to further selection. The segmentation intelligent selection method implements integrated application of multiple methods for power system security and stability assessment scenario. The method can effectively reduce the feature dimension, achieve intelligent feature selection, meets the prior knowledge of dispatchers, and facilitate the real-time monitoring of dispatching operators considering operation experience. Furthermore, it reduces the redundancy of selected feature which is conductive to the development of real-time dynamic security assessment and improves computational efficiency. The simulation result in IEEE-39 buses system verifies the effectiveness of the proposed method.