Abstract: The integrated utilization of multiple models offers technical support for substation data monitoring, demonstrating significant applied value in data analysis and fault warning systems. In this study, simulation tests and comprehensive analysis were conducted on multiple scenarios of substation data monitoring by employing Kalman filter, auto-regressive integrated moving average(ARIMA)model, Gaussian mixture model (GMM), moving average (MA) model, and systematic performance evaluation methodologies. The results show that Kalman filtering has good smoothing effect and state estimation ability in noisy observation data; ARIMA model can accurately capture long-term trend and short-term fluctuation of time series, which is suitable for load forecasting; the GMM successfully identifies low-probability anomalies through probability density analysis and achieves anomaly detection; the moving average model is capable of smoothing the data under different window sizes and analyzing short-term trends. Ultimately, system performance evaluation experiments were conducted to verify the processing capabilities in real-time monitoring scenarios, with experimental results demonstrating that high throughput and low latency are critical indicators for efficient system operation.

    Key words: substation; data monitoring; Kalman filter; auto-regressive integrated moving average model; Gaussian mixture model; moving average model; anomaly detection; system performance evaluation