Monitoring Smart Energy Systems using Multi-timescale Nexting

In today's energy grids the complexity increases and more and more autonomous control systems take over. But especially in energy systems the automatic control and decision making is critical and often human operators are required. Keeping an operator in the control-loop, good prediction algorithms can help to reduce work load and errors. Prediction algorithms supply the operator with additional information so that proactive commands can help to keep a system in an optimal state. But modeling large distributed systems is difficult. Therefore, statistical and learning based prediction algorithms are necessary to make predictions in such complex systems. In this work the Nexting algorithm, a type of temporal difference learning, is presented and applied to energy systems.     «

In today's energy grids the complexity increases and more and more autonomous control systems take over. But especially in energy systems the automatic control and decision making is critical and often human operators are required. Keeping an operator in the control-loop, good prediction algorithms can help to reduce work load and errors. Prediction algorithms supply the operator with additional information so that proactive commands can help to keep a system in an optimal state. But modeling la...     »