Embedded On-line System for Electrical Energy Measurement and Forecasting in Buildings

Abstract

Fine grained measurement of electrical energy consumption in commercial buildings is essential for improved fault diagnosis and control with impact on the overall operation as well as user comfort. An open system architecture is presented for data collection, processing and communication of measured energy patterns at the local and aggregated level. The implementation is based on an embedded development board with current and voltage sensors, supported by open-source software and packages. Suitable user and programmatic interfaces allow reliable bidirectional connection to external automation equipment and information systems. such as the Building Management Systems (BMS). Recent advances in advanced algorithms for time series pre-processing and data-driven modelling allow good quality in situ predictions for the collected measurements. A relevant example consists of neural network based learning systems which are able to provide accurate hour-ahead and day-ahead predictions that contribute to reduction of peak demand with economic and environmental impact. Integration of such platforms in higher-level Cyber-Physical Energy Systems (CPES) is further discussed.