Combining IoT and Time Series Model for Minute-Level Outlier Detection in Wind Speed Forecasting

Renewable energy optimisation and early warning systems require accurate short-term wind speed forecast. Anomalies in environmental data impair forecasting model reliability. This paper presents an integrated approach using IoT-based remote sensing and time series modelling to address the issue. IoT-based anemometer sensors collected wind speed data at one-minute intervals from December 24, 2024, to January 10, 2025. Aggregating the raw data into 5-minute intervals prepared it for the ARIMA model. This model determined temporal patterns and predicted short-term wind speeds. Analyzing residuals between observed and predicted results helped identify wind outliers. This approach is novel because it uses IoT-based continuous sensing and time series modeling for real-time environmental monitoring. Studies showed that a 65-minute frame with 5-minute intervals was best for replicating wind speed dynamics. Six cycles of outlier detection found 87 outliers. The ARIMA model improved predictions by include these outliers as exogenous variables. This emphasizes the importance of fixing time series model anomalies to improve prediction. The augmented ARIMA model with outlier corrections provides minute-level forecasts and reliable anomaly identification for renewable energy optimization and early warning systems. This study shows that new statistical methods and the Internet of Things (IoT) can improve real-time environmental and energy decisions.