The importance of Data Science and Data Engineering in Precise Fuel Management
Dover Fueling Solutions (DFS) monitors tens of thousands of underground fuel tanks and dispensers around the world each day. To offer the highest level of protection to its customers, it relies heavily on accurate data. Using both data science and data engineering to evaluate and analyze large sets of data it
collects from on-site devices, DFS is able to effectively identify fuel loss from a wide range of sources,including those caused by leaks, meter drift and fraud.
How DFS Collects Data from Fuel Stations
All raw data from on-site equipment is first gathered by the forecourt controller (FCC), which is connected to the Cloud platform. The FCC then sends this data to the Microsoft Azure IoT Hub to preprocess, and then saves it to Azure storage services, including CosmosDB and Azure Data Lakes.
Why DFS Needs to Collect Data
Running a fuel retail business is a challenging job that requires thorough monitoring and tracking of all equipment and processes on the forecourt in order to be done successfully. There are a number of very serious issues that can occur on site and lead to environmental damage and potentially severe financial implications, as well as harm to brand reputation which may affect future business. The best way to detect these problems before they become a more serious issue is by analyzing trends and patterns in data over significant time periods. For that, DFS uses carefully-engineered statistical and machine-learning models, which adapt standard techniques for industry-specific purposes by incorporating a great deal of domain knowledge from internal experts. These systems will identify instances of fuel loss accurately and will also flag up measurement and data-entry issues, which may not necessarily mean a forecourt losing fuel. In such cases, expert insight into initial algorithm outputs from skilled data engineers can help to reduce the number of false alarms. This fine-tuning also helps to improve the overall performance of DFS’ wetstock monitoring models over time, meaning it can provide a better service to its customers.
DFS systems are able to monitor and diagnose problems on numerous devices that have different operating characteristics and function in different conditions. It manages large-scale analytics and model deployment using Microsoft Azure services and infrastructure, most notably Apache Spark with Azure Databricks and Docker containers, for more complex modeling pipelines.
Data science plays an important and often crucial role in the development of DFS’ products, services and solutions. Data-driven capabilities are continually developed and refined to ensure seamless and uninterrupted operation of all fuel station equipment, which helps fuel retailers to lower operational costs and improve profit margins, whilst offering regulatory and environmental protection to keep operators legally compliant.