Research Project KonSEnz

The overall goal of the project is to address the increasingly weather-dependent energy generation processes and the associated flexible consumption with new forecasting concepts. For this subproject, the objective is to integrate continuously self-learning methods as microservices for photovoltaic and power flow forecasts into a scalable architecture. These microservices must be properly orchestrated to handle a growing number of forecast calculations and continuous optimizations. This will be demonstrated together with partners in a field test in the form of a functional prototype. The methods should allow continuous adaptation to changes in the real behavior of the forecasted units and to new units. Key features of the developed architecture include continuous learning, automated processes, robust performance with limited data availability and changes, and efficient operation. The results will also be developed with a view to future integration into the control rooms of network operators. For PV forecasting, in addition to generation depending on availability, self-consumption and associated flexibility potentials are forecasted. For vertical power flow forecasting, changes in power generation and consumption, as well as changes in the power grid, are continuously considered. The aim is to provide a toolkit for the developed methods and integrate it into a microservice architecture that combines continuous learning, transfer learning, and machine learning in operation. This will be applied to two real use cases, and the functionality will be demonstrated in a field test.