An Equivalent Consumption Minimization Strategy for a Parallel Plug-In Hybrid Electric Vehicle Based on an Environmental Perceiver

. 2022 Dec 8;22(24):9621. doi: 10.3390/s22249621

Algorithm 1 Training process of self-organizing mapping neural network (SOM)

Input:
The normalized traffic condition sample X.
Output:
The well-trained SOM network.
1:
Initializing and normalize weight W, $W_{j}, j = 1, . . ., C l a s s$ ; initial neighborhood radius $f_{n e i g h b o r} (1) = n_{0} = 3$ ; initial learning rate $f_{l e a r n} (1) = l_{0} = 0.1$ ; initial max iterations $M a x I t e r$ and iterations $I t e r$ .
2:
while $I t e r$ < $M a x I t e r$ do
3:
for i = 1 to $T \times N$ do
4:
Choose the traffic condition sample $x_{i} \in X$ .
5:
for j = 1 to $C l a s s$ do
6:
Calculate the Euclidean distance and select the nearest winning node.

$d_{i j} (x) = \sqrt{{(x_{i} - w_{j})}^{2}}$
7:
for K = 1 to $f_{n e i g h b o r} (1)$ do
8:
Update the weight of the winning node and the neighbor nodes.

$w_{k} = w_{k} + f_{l e a r n} (i t e r) \times f_{n e i g h b o r} (i t e r) \times (x_{i} - w_{k})$
9:
          Update the number of iteration, learning rate, and neighborhood radius function.

           $f_{n e i g h b o r} (I t e r + 1) = n_{0} \times exp (- I t e r / t_{1}),$

           $t_{1} = M a x I t e r / log (n_{0})$

           $f_{l e a r n} (I t e r + 1) = l_{0} \times exp (- I t e r / t_{2}),$

           $t_{2} = M a x I t e r$

           $I t e r = I t e r + 1$
10:
end for
11:
end for
12:
end for
13:
end while