交直流混联电网连锁故障特征事件智能溯源及预测方法

交直流混联电网跨区连锁故障严重威胁电网安全,调度运行面临巨大压力和事故风险.文中提出了融合知识图谱和机器学习算法的特征事件智能溯源及预测方法,实现连锁故障事故链的在线溯源和预测.将电网运行状态量和故障信息作为输入特征量,结合物理机理分析和皮尔逊系数法提取关键特征量,通过支持向量机判断特征事件间的关联关系.建立两层级的知识图谱结构,通过检测到的特征事件类型匹配上、下层事件的关联关系.根据特征事件关联关系判断结果和时序特征,基于深度优先搜索策略识别连锁故障演化路径.通过对实际电网在线数据的案例分析验证了该方法的有效性.

Intelligent Method for Characteristic Event Tracing and Prediction of Cascading Failures in AC/DC Hybrid Power Grid

Cross-region cascading failures seriously threaten safe operation of AC/DC hybrid power systems, and dispatching operation is facing tremendous pressure and accident risk. An intelligent tracing and prediction method for characteristic events is proposed, which integrates the knowledge graph and machine learning algorithm, so as to realize online tracing and prediction of cascading failure chains. The operation state and fault information of power grids are taken as input characteristic variables. Based on the physical mechanism analysis and Pearson coefficient, the key characteristic variables are extracted, and the relationship between characteristic events is judged by a support vector machine. A bi-level structure of the knowledge graph is established to match the association relationship between the upper and the lower level events based on the types of detected characteristic events. According to the judgment result of the association relationship of characteristic events and the time sequence characteristic, the evolution path of cascading failures is identified based on the depth-first searching strategy. The effectiveness of the proposed method is verified by a case study of online data from an actual power grid.