206
7 Summary and Outlook
learning-based prediction of energy consumption for hybrid simulation. The combination of hybrid simulation and machine learning is well suited for problems in
which behavioral aspects are difficult to model, but enough data of the real system
exists [Be+2019, p. 1357]. Wörrlein et al. propose the use of a hybrid system
model with a Recurrent Neural Network (RNN) encoder-decoder architecture that
returns a discrete time series when a behavior sequence has been inserted into a
neural network model of the machine. This allows the energy consumption of the
machine to be displayed for each job executed. When weighted and called, the
neural network emits the length of a job as well as a corresponding time series
indicating the quasi-continuous time consumption of the order [Wö+2019, p. 121].
While being work-in-progress, this approach could have the potential for a better
depiction of causal relationships between parameters of the simulation model and
the predicted energy consumption.
7 Summary and Outlook
learning-based prediction of energy consumption for hybrid simulation. The combination of hybrid simulation and machine learning is well suited for problems in
which behavioral aspects are difficult to model, but enough data of the real system
exists [Be+2019, p. 1357]. Wörrlein et al. propose the use of a hybrid system
model with a Recurrent Neural Network (RNN) encoder-decoder architecture that
returns a discrete time series when a behavior sequence has been inserted into a
neural network model of the machine. This allows the energy consumption of the
machine to be displayed for each job executed. When weighted and called, the
neural network emits the length of a job as well as a corresponding time series
indicating the quasi-continuous time consumption of the order [Wö+2019, p. 121].
While being work-in-progress, this approach could have the potential for a better
depiction of causal relationships between parameters of the simulation model and
the predicted energy consumption.