Carl Data Solutions provides a suite of software tools called Flow Works, that are used by municipalities to help manage their wastewater infrastructure. Their tools pull data from various sensor channels that measure variables such as water flow, velocity, and depth.
These sensors sometimes malfunction or behave unexpectedly, causing skewed readings. Since forecasting models are built on top of these sensors, skewed data can negatively influence accuracy. Currently, to account for irregularities, municipalities that use Carl Data’s flow works solutions hire consultants to manually sift through all the sensor data and modify values believed to be caused by sensor error.
Due to the overhead in time and cost, Carl Data was interested in building an anomaly detection model to automate the identification of these errors.