Project results

The project SONDER as an international project considered developing internationally compatible models and technologies.


The Austrian project share was especially devoted to interoperability. An initial IES Integration Profile was developed on managing and controlling energy storage systems using the IEC 61850 standard. The IES Technical Framework “Digitale Schnittstellen für Energiegemeinschaften in Österreich” was drafted that outlines the integration of Energy Communities in the established data exchange via the Austrian EDA platform using ebUtlilities profiles. A multi-objective control strategy and an according control module, the Community Controller, were developed and integrated in the toolbox Bifrost. Based thereon the utilisation of different flexibilities, such as energy storage systems, was evaluated in simulation studies based on real data and likely scenarios.


The Swiss project share aimed at effectively improving the accuracy of the grid energy balance prediction for subsequent optimization in several ways: (1) by validating sensors data; (2) by using novel advanced graph-based machine learning techniques to design predictors; (3) by increasing grid flexibility potential with advanced battery control algorithms based on cost optimization schemes; (4) by lifelong online adaptation for algorithms to track the grid evolution over time. Improved prediction helps to better match between forecasted and real energy balance at the local level enabling better utilisation of flexible loads, efficient energy trading at regional level and predictive grid maintenance. In addition, a sensitivity analysis was conducted on the impact of high integration levels of distributed energy resources on different low voltage distribution networks.


The Swedish project share investigated the potential of data centers operating as a prosumer and interoperability with focus on datacenters, energy grid and energy market. In detail, an edge datacentre microgrid model with controllable renewable energy resources including photovoltaics and UPS battery was developed. This simulation model was validated in both software-in-the-loop and hardware-in-the-loop, and against collected data to evaluate its accuracy. The interoperability between the standards IEC 61850, IEC 61499 and EDIEL was analyzed. An IEC 61850 model of the edge datacentre and an automatic generation framework which automatically creates an IEC 61499 automation control from IEC 61850 specifications were developed. Further, an interoperable aggregator node which translates protocols including IEC 61850, EDIEL, OCPP and openADR was created.


In international collaboration, a robust KPI framework was developed to assess the performance of smart energy systems (SES). The framework can be applied to any type of SES regardless of the application area, specifying the main SES requirements and the involved stakeholders’ objectives. Based on the results of the national project shares a benchmark bi-level framework was developed, coordinating the operation of both stationary energy storage systems and distributed energy assets. In this frame, the objectives of different stakeholders, particularly, distributed system operator (DSO), energy community (EC) and end-customers are considered. From the DSO’s point of view, the minimization of peak power costs is the main objective. As for the EC and its members, the enhancement of their social welfare is the main goal. The framework consists of two levels: (i) the upper-level controller defines the power of the stationary battery energy storage system, (ii) the lower-level controller defines the power of local battery energy storage systems, the power shedding of heat pumps (HPs), data centre (DC) and the power curtailment of photovoltaics (PVs). In terms of the upper-level controller, approaches based on Rule-based Control, Deterministic Model Predictive Control and Robust Model Predictive Control were assessed. Regarding the lower-level controller, the latter two approaches were evaluated. As for the forecasting models, Long Short-Term Memory Autoencoder, Transformer, Recurrent Neural Networks with Gate Recurrent Unit and Temporal Convolutional Networks were used for the prediction of transformer load, EC-uncontrollable load and non-EC load. Besides them, Random Forest models were applied for the prediction of PVs’ power generation and HPs’ power demand. The assessment was carried out for a typical low voltage European feeder. The study aimed to coordinate control of both centralised and distributed SES units at different grid levels, particularly, at residential and commercial buildings, EC and distribution substation level. In this manner, parts of each national project share and demo area were combined in a common use case. The evaluation of the proposed bi-level energy management framework provides directional guidelines about the approaches that need to be considered by the involved stakeholders in the future.