Research on intelligent diagnosis technology for data quality issues of power grid steady-state models based on knowledge graphs

Abstract

With the formation of the ultra-high voltage (UHV) AC/DC hybrid interconnected power grid and the large-scale integration of renewable energy, the integrated operation characteristics of the power grid have become increasingly prominent, placing higher demands on the accuracy of online analysis and decision-making. As the cornerstone of analysis and decision-making, the data quality of power grid steady-state models and real-time data directly determines the reliability of analytical results. However, current quality control of massive model data faces challenges such as low efficiency of manual inspection, reliance on experience in state estimation error analysis, and prominent data quality issues on the low-voltage side. To address these challenges, this paper proposes an intelligent diagnostic framework for data quality issues in power grid steady-state models based on knowledge graphs and machine learning. First, a knowledge graph modeling method covering power grid steady-state model maintenance and state estimation error analysis is constructed, enabling the structured representation of multi-source heterogeneous knowledge. Second, an intelligent diagnosis method based on supervised learning and reinforcement learning is proposed, achieving automatic classification, root cause tracing, and decision support for model data issues. Finally, an intelligent diagnostic software system is developed and deployed in an actual power grid dispatching and control system. Application results show that the system significantly improves the intelligence level of model data maintenance, increases the efficiency of state estimation error analysis by more than 20 times, and effectively supports the integrated online analysis and decision-making of interconnected large-scale power grids.