This thesis presents the design and implementation of SensEarth, which represents a digital twin of sensor infrastructure for acquiring, storing and analyzing data from heterogeneous web sources. The system is built as a Docker Compose stack of eight interconnected components: configurable scrapers, a middleware service (FastAPI), the main relational database CoreDB (PostgreSQL with the TimescaleDB and PostGIS extensions), MinIO object storage for raw data, a separate MonitoringDB, monitoring service (FastAPI), modeling module and a web interface (React, Vite, MapLibre). A key feature of the solution is that adding a new data source requires no programming code, only a JSON description of the target, format and value mapping. The system integrates sixteen anomaly-detection algorithms and a Prophet forecasting model, enabling measurement-quality assessment and prediction of future values. The solution prototype acquires data from ARSO stations and the Goriva.si portal and displays it on an interactive map. We demonstrate the use of the developed system through practical examples and critically compare it with the ThingsBoard and FIWARE platforms.
|
