Application of AI (ML) Tools for Model Validation and Anomaly Detection

The key factor determining the effectiveness and success of utilizing AI (ML) technologies is the quality of the data applied. The utilization of inaccurate data can result in the unreliable and incorrect outcomes for systems and applications powered by AI

(ML). This, in turn, can result in incorrect decision-making and subsequent negative consequences.

Traditionally, the validation of data and anomalies detection in energy sector have relied on formalized rules and validation algorithms based on mathematical statistics methods,

Kirchhoff laws and other pre-established criteria, including mathematical apparatus utilized for the execution of state estimation calculations. However, these rules and algorithms cannot be comprehensive and, in some cases their implementation might be difficult, such as when analyzing insulation measurements or other non-electrical parameters.

The article describes the issues of data preparation for the operation of applications and decision support tools and techniques in managing the increasing complexity of power grids. The modern approach to data validation and anomaly detection is proposed, involving the use of AI (ML) technologies. Application examples are provided and the effectiveness of this method in comparison with traditional approaches is analyzed.

The study demonstrated that the most effective is a hybrid approach, combining traditional validation and anomaly detection methods with methods based on AI (ML) technologies.