Control-Center analysis of fault events and load forecasting for 115/22 kV transformers using outlier filtering and Gradient-Based Holt–Winters parameter tuning

• Automated fault-event analysis: PyAuto3 provides an automated workflow for analyzing 115/22 kV transformer trip events. By parsing control alarms in real time, it identifies and verifies trip causes without relying solely on operator expertise. • Fault Tree Analysis integration: PyAuto3 integrates Fault Tree Analysis with the power-mismatch-at-bus principle to validate events and distinguish breaker-failure relay operations from normal trips. • Optimized load forecasting: For load prediction, PyAuto3 applies Holt-Winters exponential smoothing, with parameters α, β, and γ tuned using the limited-memory BFGS optimization method. • Operational efficiency and validation: In testing across five transformer trip events, PyAuto3 reduced analysis time from an average of 448 seconds to 35.8 seconds while maintaining 100% detection accuracy. Rapid and accurate diagnosis is essential to stop cascading outages when transformers in transmission networks trip unexpectedly. Control center operators are required to rapidly process numerous SCADA alarms. However, conventional manual analysis is both time-consuming and susceptible to errors. This study presents PyAuto3, an automated workflow that combines fault tree analysis with bus power-mismatch verification to streamline event screening. The system employs three complementary outlier detection methods—isolation forest, rule-based thresholds, and interquartile range—in conjunction with three correction techniques, namely, PCHIP interpolation, rolling-mean adjustment, and linear patching. The L-BFGS algorithm optimizes the parameters α , β , and γ in Holt–Winters exponential smoothing for an enhanced load forecasting accuracy. Validation with five detected 115/22 kV transformer trips from 2023 to 2024 revealed that the system could detect all of them correctly and reduced the time required to evaluate these from 448 s to 35.8 s (a 92 % improvement). After repairing the outliers, the signal-to-noise ratio increased from 11.44 dB to 28.38 dB, which considerably enhanced the quality of the data. PyAuto3′s graphical outputs assist operators in identifying overload risks when the outage load is transferred to a different substation. This helps these make smart decisions regarding transferring load. PyAuto3 improves the situational awareness and response speed in utility control centers by making these less dependent on operator expertise and less likely to commit mistakes. Although the method was developed for Thailand’s 115/22 kV interface, it may be used directly on worldwide systems that have similar SCADA quality and operational problems. Additionally, it lays the groundwork for future improvements to more complex substation topologies and adaptive forecasting models.