Design of an Automatic Control Algorithm for Energy-Efficient ...
Design of an Automatic Control Algorithm for Energy-Efficient ...
Design of an Automatic Control Algorithm for Energy-Efficient ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
6 The optimiser: <strong>an</strong> evolutionary algorithm approach 56<br />
• Select the control values with respect to the last one applied to avoid jumps.<br />
• Prevent control outputs which exceed limits by replacing them with p<strong>an</strong>ic h<strong>an</strong>dlers,<br />
i.e. special outputs.<br />
A flowchart <strong>of</strong> the algorithm is found in Figure 6.3. The methods applied are<br />
described in detail in the next sections.<br />
����<br />
����������������<br />
���������<br />
������������<br />
�����������<br />
���������<br />
����������<br />
����<br />
������������<br />
�����<br />
������<br />
�������<br />
��������������<br />
����������������<br />
������������<br />
�������������<br />
��������<br />
�����������������������������������<br />
������������<br />
���������<br />
�������<br />
�������������<br />
�������<br />
����������<br />
�������� ��������� ����������<br />
���������� ��������<br />
����������������<br />
���������������<br />
���������<br />
����������<br />
����������<br />
���������������<br />
��������������<br />
����<br />
������������<br />
�����<br />
����������<br />
����������<br />
������������<br />
�������������<br />
�����������<br />
������������<br />
�����������<br />
�����������<br />
�������������<br />
����������<br />
���<br />
����<br />
������������<br />
��������<br />
Figure 6.3: Flowchart <strong>of</strong> the incremental multi-objective evolutionary algorithm.<br />
At the start <strong>of</strong> each time-step the system states (temperatures, humidity <strong>an</strong>d other<br />
sensor values) are updated. They are then used in the system prediction. The population<br />
is built up by the elite from the last run <strong>an</strong>d a const<strong>an</strong>t base population. The remaining<br />
individuals <strong>for</strong> the population are r<strong>an</strong>domly chosen from the design space. Using the<br />
system model, the future states are predicted <strong>for</strong> each individual. Based on these the<br />
objectives are evaluated <strong>an</strong>d the results stored. In the following step the population is<br />
compared using the limits <strong>an</strong>d priorities r<strong>an</strong>king scheme.<br />
The r<strong>an</strong>king is decisive <strong>for</strong> the selection as parent. The parent couples are recom-<br />
bined <strong>an</strong>d the resulting <strong>of</strong>fspring mutated. The objectives are evaluated <strong>for</strong> the children<br />
<strong>an</strong>d a r<strong>an</strong>king together with the parent generation is conducted. Finally this r<strong>an</strong>ked list<br />
��