Brad Lehman,

Northeastern University, Boston, MA lehman@ece.neu.edu

Seminar at IEEE COMPEL; June 26, 2018

Smart Solar Energy for the Smart Grid

Outline 1. PV Basics and Problem Statement

2. Advances in Smart Solar PV’s: a) Smart Fault Detection

b) Dynamic Reconfigurability

c) Lego Solar Panels

d) Modular Differential Power Processing

3. Conclusions

• Presenting research, with gratitude from many students and colleagues: D.

Nguyen, Y. Zhao, P. Li, Y. Li, J. Huang, F. Khan, R. Balog,, S. Lu, S. Shu, L. Yuan, G. Spagnuolo, G. Petrone, C. Ramos Paja, Y. Zheng, B. Scandrett, Y. Zheng, C. Liu, D. Li, J.-F. De Palma, K. Kim, P. Krein, R. Pilawa-Podgurski, B. Scandrett, M. Gutierrez, and others…

• Funding from NSF, PowerFilm, Mersen, DOE (SunShot), US Army Natick, DOD, Arpa-e

2 2

3

What do you call a bear that uses renewable

energy?

4

1. PV Basics and Problem Statement

Central Inverter: PV, power conversion unit, protection devices and utility grid.

5

6

Different approaches grid-connected inverters

a. Centralized technology.

b. String technology.

c. Multi-string technology.

d. AC-module

Optional

(DC optimizer technique not shown) Figure: S. B. Kjaer, J. K. Pedersen and F.

Blaabjerg, IEEE Transactions on Industry

Applications, Sept.-Oct. 2005

3

PV1

PV2

PV3

PVn

DC/DC

DC/DC

Voltage Bus

DC/DC

DC/DC

DPP converter

Newer Approach: Differential Power Processing

Work by K. Kim, R. Pilawa-

Podgurski, P. Krein, R. Balog,

S. Qin, P. Shanoyand, D.

Maksimovic, and others…

8

Do you agree? • Many of the solar MPPT methods were designed 40

years ago and have not taken advantage of modern technological advances: • Computer/microprocessor revolution • Machine learning • Wireless communication

• PV installation hardwired and left alone until failure. What about expansion of the system?

How can we take advantage of the advances we have seen to create new approaches in PV systems??

2. Advances in Smart Solar Energy

9

10

Fault detection

Fire hazards in a 1,208kW PV array, in Mount Holly,

North Carolina, in 2011.

Traditional Overcurrent Protection Devices (fuses) have “Blind Spots” in PV Systems with Centralized Inverters!!

2. A) PROBLEM - FAULTS

11

Sometimes the MPPT makes it worse: moves operating point to lower current before the fuses melt

vf

Utility

grid

Centralized inverter

IPV

VPV

PV module

+

_

OCPD (Series fuse for

overcurrent protection)

String 1 String 2 String 6 String 12

Pos.

Neg.

I1 I6 I12

Line-line faults

created by a switch

LL-m

LL-2

LL-1

String 7

I7

PV

monitoring

board#1

I2

+

_PV

monitoring

board#2

Line-line fault with

m number of

modules mismatch

Current sensor

Switches to

create faults

PV module #1

PV module #2

PV module #3

PV module #4

PV module #9

PV module #10

PV module #11

PV module #12

...

More LL

faults

Problem: Some Line-Line faults or night time faults cannot be cleared.

Zhao, Y. de Palma, J.-F, Mosesian, J., Lyons, R, and Lehman, B., IEEE

Transactions on Industrial Electronics,September 2013.

(See also work by F. Khan, J. Johnson,

X. Wang, A. Etemadi, and others)

12

Normal

String PV

Current

Fault at t= T1 may be current

limited in PV (low irradiance,

line-line)

Oh No! MPPT reduced fault current so fast that fuses cannot clear fault!

13

Machine Learning in PV Strings

Compare string currents to find faults

Fault detection using outlier rules at PC (Matlab GUI)

Real-time operation

Combiner box (Mersen): DSP fuses, Machine Learning

Statistic outlier detection rules

• Univariate variable {xi}, i = 1….n, x = string current

• xi is an outlier if

• Reference value x0, a measure of variation and a threshold

• Upper bound , lower bound

𝛼

Reference: R. K. Pearson, Mining imperfect data: Dealing with contamination and incomplete records, Society for Industrial and Applied

Mathematics, 2005

xi

p(x)

x0

Outlier?

An outlier

x1 x2

True distribution

14

Zhao Y, Ball R, Mosesian J, de Palma JF, Lehman B., IEEE Transactions on Power Electronics, May 2015.

Statistic outlier detection rules

• Three sigma rule

– Might be the best-known criterion, but poor performance in practice

– Breaks down if CL >10%

• Hampel identifier

– Breaks down if CL >50%

• Boxplot rule

– Breaks down if CL >25%

• Outlier rules are updated at each sampling time.

xL is lower quartile (25th percentile)

xU is the upper quartile (75th percentile)

ˆ3ˆ| |ix

Sample mean

Sample variance

15

Photo of the experimental setup > Fault algorithm is running in a GUI at PC

Matlab GUI

DC electronic load

DC power supply

GreenString in

the combiner box

Power resistors

16

Video Demo

17

2. b) Dynamic Reconfiguration

18

Why keep all the wiring connections the same between the PV panels all the time?

SMART PV PANEL

• Smart Solar Arrays

– Merge electronics to be inside the solar

panel instead of in a separate room

– Automatically adjusts to shadows and

the environment by changing its

interconnections

– Senses, diagnoses, and alerts user of

problems with the array (cracks,

malfunctions, etc.)

– Heals itself when solar cells degrade

– Smart grid interaction

– Optimum power output at all times

Lab Prototype: Smart Solar Array

The solar

fixed part

PV1 PVA1

V1

IOUT

IA1

VO

UT

Temp, IOUT

PV2

IF2

IA2

PVA2

IA(m-1)IOUT

PVA(m-1)

IAm

PVAm

PVm-1

PVm

The adaptive

bank

A/D CONVERTER

INTERFACE

R1

R2

R(m-1)

C1

C2

Cm

The

Switching

Matrix

Rm

C(m-1)

Vbatt.

V2

Vm-1

Vm

VOCA1

VOCA2

VOCA(m-1)

VOCAm

{VOCA1,VOCA2,…,VOCAi,…,VOCAm}{V1,V2,…, VAi,…,VAm}

r

D. Nguyen and B. Lehman, IEEE Trans. on Industrial Electronics, 2008

19

• Place switches between solar PV cells, strings, or arrays

• Switch between series and parallel connections in

response to clouds or shadows

• Increase the power output of the solar PV array

D. Nguyen and B. Lehman, (2008), Review paper G. Spagnuolo, IEEE Industrial Electronics Magazine, 2015

• Fixed solar cells: m x n

When PV cells in fixed part

are shaded or blocked, the

PV cells in reconfigurable

part will rearrange and

connect to shaded rows.

Shaded cell

• Add m x 1 reconfigurable

solar cells

Slow Reconfigurable Panels

20 20

Even Without Faults, Are There Benefits to Changing PV Connections?

21

21

2. c) Lego Solar Panels

“Solar Blanket”-Modularity

22

Y. Zheng, Y. Li, S. Sheng, B. Scandrett and B. Lehman (APEC 2016, 2017),

Introduction – modular solar panels

New challenges:

• Need to carefully design control to Shed power if producing too much power for load.

Turn off MPPT if user connects to another MPPT

Submodule 1

DC-DC

DC-DC

Power management/ charger

Battery

Loads

Subpanel

PV Bus

Submodule n

P

23

System Operating Mode

Scenario I: Resistive load,

PMPPT Pload_max

IO

VH

VL

Rload

PPV = PMPPT Pload_max

OL

OH

O

O H

O L

VPVBus

Scenario II: Resistive load,

PMPPT > Pload_max

I

VPVBus

O

VH

VL

Rload

Shed Power

PPV = P1oad_maxPMPPT

OH

OL

Scenario III: Constant power

load, PMPPT Pload_max

IO

VH

VL

OL

OH

VPVBus

PPV = PMPPTPload_max

Scenario IV: Constant power

load, PMPPT > Pload_max

IO

VH

VL

Shed Power

OH

OL

VPVBus

PPV = P1oad_max

PMPPT

24

Y. Li, et al APEC 2016

System Operating Mode

Scenario I: Resistive load, PMPPT Pload_max

IO

VH

VL

Rload

PPV = PMPPT Pload_max

OL

OH

O

O H

O L

VPVBus

• All PV subpanels are in individual MPPT mode

• The operating point would be the intersection (O) of load curve

Rload and power contour PMPPT

• The operating point moves between OL and OH along the red

straight line (solid), or between O’L and O’H along the blue curve

with source/load changes

25

System Operating Mode

Scenario II: Resistive load, PMPPT > Pload_max

I

VPVBus

O

VH

VL

Rload

Shed Power

PPV = P1oad_maxPMPPT

OH

OL

• Controller must shed power

• Option 1: Do NOT have all modules in MPPT mode

• Option 2: turn off power from individual PV module.

• The area shaded in the Figure represents the amount of shed power

When PPV is too high, the voltage might rise above threshold VH

26

PV1

PV2

PV3

PVn

DC/DC

DC/DC

Voltage Bus

DC/DC

DC/DCDPP converter

One Last Idea: What happens when we combine Lego (Modular) solar panels

with Differential Power Processing

27

2. d) Modular Differential Power Processing (mDPP)

Voltage Bus

PVm,1

PVm,2

PVm,3

PVm,n

DC/DC

DC/DC

DC/DC

DC/DC

Controllable current source

PV1,1

PV1,2

PV1,3

PV1,n

DC/DC

DC/DC

DC/DC

DC/DC

Iout is current difference between PV panels

Virtual PV panel

Vcap is voltage difference between the string and the bus

PVVmpp=?VImpp=0A

28

C. Liu, D. Li, Y. Zheng and B. Lehman (2017, COMPEL) and (2018, APEC)

29

Advantage of modularity

7V mDPP

Plug out

Advantage of modularity

8V

30

7V mDPP

Plug in

8V

31

Advantage of modularity

Simulation results

3 PV panel in series

2 PV string in parallel

PV13 - voltage mismatch

PV 22 - current mismatch

PV#11 PV#12 PV#13 PV#21 PV#22 PV#23 Total

Ideal MPP (P) 13.23 12.3 10 12.59 6.13 12.81 67.06

Simulated Power with mDPP 13.23 12.3 10 12.58 6.13 12.81 67.05

32

Simulation results • Comparison between Different Methods

TABLE II. SYSTEM EFFICIENCY STUDY OF DIFFERENT DPP STRUCTURES

cMPPT dMPPT sDPP sDPPcc mDPP(ours)

Output Power (Pout - Watts) 53.56 62.36 42.60 61.11 65.86

System Efficiency (ηsys - %) 79.86% 92.99% 63.51% 91.11% 98.21%

cMPPT: centralized MPPT converter- one converter for all PV panels dMPPT: distributed MPPT converter- one converter for each PV panel [1] sDPP: series DPP method [2] sDPPcc: series DPP converter with centralized converter [3]

[1] Y. Zheng, Y. Li, S. Sheng, B. Scandrett, and B. Lehman, "Distributed control for modular plug-and-play subpanel photovoltaic converter system,"

in 2017 IEEE Applied Power Electronics Conference and Exposition (APEC)

[2] S. Qin, S. T. Cady, A. D. Domínguez-García, and R. C. N. PilawaPodgurski, "A Distributed Approach to Maximum Power Point Tracking for

Photovoltaic Submodule Differential Power Processing," IEEE Transactions on Power Electronics

[3] S. Qin, C. B. Barth, and R. C. N. Pilawa-Podgurski, "Enhancing Microinverter Energy Capture With Submodule Differential Power Processing,"

IEEE Transactions on Power Electronics

6

,max,

1

out outsys

ideali mpp

i

P P

PP

Harvest most power 98.21%

33

Experimental results

Up to 97.6% efficiency with 90% efficiency converter

34

CONCLUSIONS

1. Legacy approaches to operating PV arrays?

2. Hidden Faults in PV can now be discovered

3. Dynamic Reconfiguration is possible

4. Modular solar panels due to miniaturization

and control in power electronics

5. Differential Power Processing might

improve performance of Lego PVs

35