Networked Embedded Systems Lab

Prof. Marco Zimmerling

Introduction to Embedded Systems – WS 2022/23

Exercise 6: Low Power II

Task 1: Energy Harvesting

Consider a processor with negligible leakage power dissipation and a dynamic power dissipation that depends
as follows on the operating frequency f in Hz:
 3
f
Pdynamic (f ) = mW
1 MHz

Whenever the processor is idle it enters the zero power state without additional overhead. The set of hard
real-time tasks τi listed in Table 1 shall be executed on the processor.

Task τ1 τ2 τ3
Arrival Time [ms] 0 0 0
Period [ms] 6 4 12
Relative Deadline [ms] 6 4 12
Cycles [×103 ] 2 1 2

Table 1: Characteristics of the set of hard real-time tasks τi to be executed.

The system has a battery with an energy level Ebat (t) and it is replenished by a constant power source with
P (t) = Pin .

1. Assume an initial battery charge of Ebat (t = 0) = 6 µJ, a constant input power Pin = 0.5 µJ/ms, and a
constant clock frequency of f = 1 MHz for task processing. Schedule the tasks τ1 , . . . , τ3 according to
the Earliest-Deadline-First (EDF) algorithm and draw the evolution of the battery energy level Ebat (t)
in Figure 1 for the time interval [0 ms, 12 ms].

2. Assume the battery does not run out of charge. Prove or disprove the following statement:
To maximize the energy stored in the battery at the end of each hyper-period (12 ms), all tasks τi
have to be executed at the same frequency and this frequency leads to a utilization of 1.0.
Hint: either provide the main arguments or a formal proof, both are accepted.

Task 2: Solar Cell Characteristics and Maximum Power Point Tracking

In this task we consider a solar energy harvesting system that performs maximum power point tracking for
optimal harvesting. Specifically, the system employs the power point tracking algorithm introduced in the
lecture. The algorithm dynamically adapts the operating point in discrete voltage steps ∆ to match the

1
 10

Battery charge state Ebat (t) [µJ]

9
8
7
6
5
4
3
2
1
0
0 1 2 3 4 5 6 7 8 9 10 11 12
Time [ms]

Figure 1: Evolution of the battery charge state Ebat (t).

maximum power point P ∗ as close as possible. The flow chart describing the algorithm is shown in Figure 2.
The solar cell connected to the systems has an I-V -characteristic that is described by the following model:
   
V
I(V ) = G · 1 A − exp − 1 · 0.01 mA, (1)
0.1 V

where I is the solar cell’s output current, V its output voltage and G is the relative solar irradiance (unit-less
with 0 ≤ G ≤ 1).

start new iteration k := k+1

sense V[k], I[k]

P[k] = V[k] · I[k]

yes no
P[k] > P[k-1]

yes no no yes
V[k] > V[k-1] V[k] > V[k-1]

set V[k+1] = V[k] + Δ set V[k+1] = V[k] - Δ

end iteration k

Figure 2: Flow chart describing the Maximum Power Point Tracking algorithm.

1. Compute the power extracted from the solar cell when operating at solar cell voltage V = 0.7 V for
relative irradiances of G = {0.1, 0.2, 0.5, 1.0}.

2. Execute the power point tracking algorithm by hand, once for G = 0.1 and once for G = 1.0. Use
a step-size ∆ = 0.05 V for the voltage adjustments and start with iteration k = 1, V [0] = 0.7 V and
V [1] = 0.75 V. Based on these result, determine a lower bound on the maximum power point P ∗ .

3. Consider now a photovoltaic panel (PV) consisting of two identical cells arranged in series, as illustrated
in Figure 3. Due to partial shading of the PV panel, cell 1 receives the full relative irradiance of G1 = 1.0,
while cell 2 receives only G2 = 0.1. Find a reasonable upper bound on the power that can be generated

2
 using this PV panel. How does the extracted power compare to cell 1, if it is used standalone with the
full relative irradiance of G1 = 1.0?

V1 V2
+ - + -

IPV
VPV
Figure 3: A photovoltaic panel consisting of two cells connected in series.

Task 3: Application Control

We consider an application control scenario with the harvested energy in time interval [t, t + 1] of p(t), used
energy of u(t), battery capacity B and battery charge level b(t) ∈ [0, B]. The utility function used for
√
application control is defined as µ(u) = u. Where not specified otherwise, the energy values of p(t), u(t),
b(t), and B are given in watt-hours (Wh) and the time t is given in hours (h).
Energy harvested p(t) [Wh]

b
7
b b
6
b b
5
b b
4
b b
3
b b
2
b b
1
b b b b b b b b b b b
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Time t [h]

Figure 4: Energy harvesting profile p(t) for a daily cycle. The values of p(t) are all integers.

1. Consider the energy harvesting profile p(t) given in Figure 4 that repeats daily. What is the maximum
average power udmax that can be used by the system?

2. Given the knowledge of the daily energy input profile p(t) in Figure 4, calculate the minimal battery size
Bmin such that the used energy satisfies u(t) = 2 for every time interval during a day. Complete the
diagram in Figure 5 with the daily evolution of the used energy u(t) and the battery charge state b(t)
at the beginning of the interval for the found battery size Bmin .

3
 10

Energy use u(t) [Wh]

8

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Time t [h]
25
Battery fill level b(t) [Wh]

20

15

10

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Time t [h]

Figure 5: Energy harvesting profile p(t) for a daily cycle. The values of p(t) are all integers.

In the following tasks we assume a system with a battery size that is limited to B = 11 Wh. For application
control we assume an observation interval of T = 24 h and require that the current battery level is reached at
the end of this observation interval, i.e., b(t + T ) = b(t) ∀t ≥ 0.

3. Determine an optimal energy usage function u∗ (t) that maximizes the minimal used energy and maxi-
mizes the total utility.

4. Unexpectedly, we do not harvest any energy, i.e. p(t) = 0, during one of the intervals t ∈ [6, 18].
The exact interval during which no energy is harvested is unknown. The energy use is still u∗ (t) as
computed above. Explain the consequences, considering all possible scenarios. Are there any system
failures happening and at what time of the day?

5. In this task we use a finite horizon control scheme for application control. Suppose that we unexpectedly
have zero energy in interval t = 12, i.e., p(12) = 0. What is the computed energy use in this case?