Example: Using the previous settings of 3 five minute intervals, and a new setting of 120%
prediction factor, the working of the Predictive Window Demand could be described as follows:
At 12:10, we have the average of the subintervals from 11:55-12:00, 12:00-12:05 and 12:05-12:10.
In five minutes, we will have an average of the subintervals 12:00-12:05 and 12:05-12:10 (which we
know) and 12:10-12:15, which we do not yet know. As a guess, we will use the last subinterval
(12:05-12:10), as an approximation for the next subinterval (12:10-12:15). As a further refinement,
we will assume that the next subinterval might have a higher average (120%) than the last
subinterval. As we progress into the subinterval, (for example, up to 12:11), the Predictive Window
Demand will be the average of the first two subintervals (12:00-12:05, 12:05-12:10), the actual
values of the current subinterval (12:10-12:11) and the prediction for the remainder of the
subinterval, 4/5 of 120% of the 12:05-12:10 subinterval.
# of Subintervals = n
Subinterval Length = Len
Partial Subinterval Length = Cnt
Prediction Factor = Pct
Sub n
Len
Len
−
1
∑
Value
i
Sub
=
i
=
0
Len
Cnt
−
1
∑
Value
i
Partial
=
i
=
0
Cnt
⎡
n
−
2
∑
Value
⎢
⎢
Partial
+
i
=
0
⎢
⎢
⎣
⎡
n
−
2
∑
Sub
⎢
i
Sub
⎢
+
i
+
=
0
n
−
1
⎢
⎢
⎣
Electro Industries/GaugeTech
Sub 1
...
Len
⎤
⎥ ⎡
i
⎡
Len Cnt
−
⎡
⎥
+
1
−
⎢
⎢
⎢
n
Len
⎥
⎣
⎣
⎣
⎥
⎦
⎤
⎥ ⎡
Sub
Len Cnt
−
⎡
⎥
n
x
0
−
1
⎢
⎢
x n
−
Len
2 (
1)
⎥ ⎣
⎣
⎥
⎦
Doc # E151701
Sub 0
Partial
Len
Cnt
⎤
⎤
⎤
xPct
⎥
⎥
⎥
⎦
⎦
⎦
⎤
−
⎤
xPct
⎥
⎥
⎦
⎦
Predict
Len
13-7