The project has developed a community-based aggregator function for the smart grid. For such a system there is a need for robust and secure communication interfaces and protocols, efficient and secure data acquisition and storage, and mechanisms to provide service guaranties.
SEMIAH concentrates on software architecture and components, algorithms, service descriptions, as well as interoperability for an aggregator function and we have developed a prototype to be used in the project.
The control of the appliances is based on RESTful architectural principles.
In order to efficiently match the electrical consumption in households with the generation of electricity produced from RES, it is necessary to aggregate a large number of loads according to users’ flexibilities. SEMIAH has developed semantic technologies for the aggregation of data based on a common ontology to handle data integration from various sources.
The semantic model developed in SEMIAH provides an improved way to manage metadata of a large number of devices and gives more flexibility to the system to deal with new devices.
To ensure the integration of large GW of RES while maintaining the stability of the grid, it is necessary to use accurate and efficient forecast techniques for the energy production and consumption. Thus, the DR function developed in SEMIAH makes use of cutting-edge algorithms developed at FRAUNHOFER to forecast wind energy supply and PV production at specific sites.
Moreover, a load estimator, which produces probabilistic load forecasts for the households, has been developed.
The grid stability is being assessed by means of local metering data combined with adequate models for the prediction of the impact of local production on the associated low- and medium voltage grid segments. The planning and scheduling of electrical loads are based on these forecasting algorithms and grid stability models and stochastic optimization are used in the scheduling algorithms. Restrictions from DSO grid management and market energy prices are also taken into account.
Overall, the project has aimed to design a scalable solution that is able to manage entire systems and millions of appliances and TWhs of energy by designing an architecture that is inherently parallelisable, scalable, reliable and robust.
Furthermore, the system has real-time Demand-Response within less than 5 minutes response time, making it competitive against other projects. This is also a good foundation for investigating parallelisable algorithms that have a reasonable computational efficiency.
Additionally, the semantic model supports reasoning and can furthermore be augmented using Artificial Intelligence methods – such as reinforcement learning, neural networks or clustering – to improve the energy scheduling from both a quality-of-service perspective, to keep the voltage and frequency stable and – from an economic perspective – to be as profitable as possible.