Outliers and irregular patterns can undermine forecasting and analysis of sensor data. Two approaches—autoregressive cost updates and wavelet-based detection—offer different ways to identify anomalies, with the choice depending partly on whether disruptions are isolated or consecutive.