6.3 Optimization Experiments
181
results of the optimizations were very close and that the use of the high-resolution
data proved to be rather disadvantageous, due to significantly longer simulation
runtimes times. For the calculation of the mean values, all energy values per state
are summed up and divided by the total number of energy values. Although the
energy consumption visualization in the simulation shows significantly less volatile curves due to averaging, the machine state changes can be seen clearly in
the visualization of the simulation (Figure 6.21). Extreme values are no longer
displayed.
The execution of the total consumption optimizer experiment revealed nearly
identical parameter configurations (compared to the experiment run with load profile data) for the idle and standby optimizer parameters as optimal (Figure 6.22).
The minimum deviations of 0.15% for the standby optimizer parameter and
0.33% for the idle optimizer parameter can be justified by rounding errors in
the averaging calculations.
Following the determination of the optimal parameter set to fulfill the objective function for the total energy consumption, the peak optimization experiment
was started. Within the scope of the optimization, however, it was not possible to
determine an optimal offset parameter that achieves a reduction of the occurring
peak loads in the mean value scenario within the permitted limits for the parameter. Looking at the data of the first few minutes of the simulation, it quickly
becomes clear why the load peaks cannot be reduced. The offset parameter can
cause delays of up to five minutes.
During the warmup phase, the maximum peak of the mean value scenario
is reached for the first time. The entire warmup has a duration of 20 minutes
and the peak occurs over the entire duration of the warmup phase. An offset of
five minutes cannot counteract here. A reduction of the power peak load through
the parameter variation in the peak optimization experiment run is therefore not
meaningful (Figure 6.23).
Comparing the mean value scenario with the optimization scenario (1-min
resolution), it becomes clear that only very minor deviations of 0,4% in the depiction of the total energy consumption for the same production output quantity can
be noted (Table 6.7). The depiction of the exact consumption peaks is not possible in the mean value scenario. While it is possible to optimize the total energy
consumption by using the total consumption optimizer experiment, the peak load
optimization with averages does not lead to useful results. This is solely due to
the fact that extreme values are no longer recognizable in the consumption profiles due to averaging and is neither caused by failure of the methodology nor the
simulator.
181
results of the optimizations were very close and that the use of the high-resolution
data proved to be rather disadvantageous, due to significantly longer simulation
runtimes times. For the calculation of the mean values, all energy values per state
are summed up and divided by the total number of energy values. Although the
energy consumption visualization in the simulation shows significantly less volatile curves due to averaging, the machine state changes can be seen clearly in
the visualization of the simulation (Figure 6.21). Extreme values are no longer
displayed.
The execution of the total consumption optimizer experiment revealed nearly
identical parameter configurations (compared to the experiment run with load profile data) for the idle and standby optimizer parameters as optimal (Figure 6.22).
The minimum deviations of 0.15% for the standby optimizer parameter and
0.33% for the idle optimizer parameter can be justified by rounding errors in
the averaging calculations.
Following the determination of the optimal parameter set to fulfill the objective function for the total energy consumption, the peak optimization experiment
was started. Within the scope of the optimization, however, it was not possible to
determine an optimal offset parameter that achieves a reduction of the occurring
peak loads in the mean value scenario within the permitted limits for the parameter. Looking at the data of the first few minutes of the simulation, it quickly
becomes clear why the load peaks cannot be reduced. The offset parameter can
cause delays of up to five minutes.
During the warmup phase, the maximum peak of the mean value scenario
is reached for the first time. The entire warmup has a duration of 20 minutes
and the peak occurs over the entire duration of the warmup phase. An offset of
five minutes cannot counteract here. A reduction of the power peak load through
the parameter variation in the peak optimization experiment run is therefore not
meaningful (Figure 6.23).
Comparing the mean value scenario with the optimization scenario (1-min
resolution), it becomes clear that only very minor deviations of 0,4% in the depiction of the total energy consumption for the same production output quantity can
be noted (Table 6.7). The depiction of the exact consumption peaks is not possible in the mean value scenario. While it is possible to optimize the total energy
consumption by using the total consumption optimizer experiment, the peak load
optimization with averages does not lead to useful results. This is solely due to
the fact that extreme values are no longer recognizable in the consumption profiles due to averaging and is neither caused by failure of the methodology nor the
simulator.