University of Oxford

Power Electronics for Grid Scale Energy Storage

Getting the most out of your cells

Dr Dan Rogers

Senior Research Fellow, Department of Engineering Science

dan.rogers@eng.ox.ac.uk

UKES 2016, Birmingham

1st December 2016

Overview

What is power electronics?

A quick look at how today’s grid-scale storage systems are normally connected to the grid

How ‘more’ power electronics can improve

Reliability

Accessible system capacity

The challenges of control and management of very large numbers of cells

Some hardware and results

2

What is power electronics?

“The use of power semiconductor devices to control and convert electrical energy”

3

DC

sourceInput filter Power transistors Output filter AC source

Power electronics for grid storage

Suitable for relatively small systems: essentially a ‘scaled up EV’

2.5 V, 20 Ah cells (50 Wh): A 1 MW / 1MWh system needs 20,000 cells

If pack voltage is 500 V then pack is 200s100p and DC bus current is 2 kA

10 cm2 cross section copper wiring and large 𝐼2𝑅 losses throughout the system

Monolithic systems of this scale start to become impractical4

Battery pack DC-DC

converter

(optional)DC-AC

converter Step up

transformer Grid

300-600 Vdc

700 Vdc 400 Vac11 kVac

Statistical modelling of packs

Notation

𝑁 = number of cells in system

𝑐𝑖 = capacity of cell 𝑖

𝐶 = capacity of system

Ideally we would like

𝐶 = 𝑖=1𝑁 𝑐𝑖

But for a naïve series string design

𝐶 = 𝑁min𝑖=1𝑁 𝑐𝑖

𝑐𝑖 varies from cell to cell and from moment to moment*

𝑐𝑖 𝑡 = 𝑐𝑖 0 − 𝑑𝑖𝑡 𝑒𝑖 𝑡

5

initial capacity degradation rate

(𝑑𝑖 > 0)

catastrophic failure

(1 → 0 at some 𝑡)

*modelling framework developed in collaboration with Matthias Troffaes (Durham) and Louis Aslett (Oxford)

𝑐𝑖, 𝑑𝑖 and 𝑒𝑖 vary from cell-to-cell:

we assume normal distributions or

parameters in order to model large systems

The problem with large series packs

Cell capacities are fairly tightly distributed at SoL

95% of cells within ±0.1 of mean

Cell degradation is relatively ‘slow’

10% decrease in mean capacity

over lifetime

~90% have more than 0.8

capacity left at EoL

Even at SoL, only ~80% of total cell capacity is available to the system

It’s quite likely we will have only 60% of capacity accessible at EoL

6

start: 𝑐 0

end: 𝑐 EoL

end: 𝐶 EoLstart: 𝐶 0

Packing cells into modules

A solution to this problem is to break the system up into small modules that are managed individually

7

𝑁 modules in

the system

𝑀 cells in

a moduleSystem contains 𝑀𝑁 cells

But module voltage is only 𝑀𝑉𝑐𝑒𝑙𝑙∴ modules should be connected in series

Module power electronics

provides control over the average

current flowing in each module

The ‘depth’ of modularisation strongly influences lifetime behaviour

If 𝑀 = 1 we achieve complete capacity utilisation

Choosing 𝑀 = 5 gives ~20% gain in EoL capacity

Modularisation gives more predictable EoL capacity

This assumes no ‘random complete failure’ mechanism!

Benefits of modularising a system

8

From no

modularisation

to using 𝑀 = 5modules

Start of life

End of life

More benefits of modularisation

Adding random failure makes modularisation even more attractive

Here, probability of cell failure is

0.001

i.e. we expect about 20 cells to fail

over the lifetime of the 20,000 cell

system

Of course, in a real system, maintenance will replace failed cells

But system must not suffer downtime as a result of a single cell (or power electronics) failure

Modularisation of some sort is

required

9

𝑀 > 500 produces

very unpredictable

system EoL capacity

End of life

Start of life

Option 1

10

Power block 1

Power block 2

Power block M

Module

1.1

Module

2.1

Module

Ns.1

Module

1.Np

Module

2.Np

Module

Ns.Np

Grid

Step-up

transformer

Conventional battery pack

Cell 1

Cell 2

Cell Nc

Module

Option 2

11

Module

1.1

Module

2.1

Module

Ns.1

Module

1.Np

Module

2.Np

Module

Ns.Np

Intelligent battery pack

Grid

Medium-

voltage

converter

Option 3

12

Phase B Phase C

Cascaded H-BridgeGrid

Module

1

Module

2

Module

Nhb

N

Phase A

Markov models for reliability modelling

Model a system as composed of many components

The model includes all possible system states coupled by transitions that occur as components fail

Chain terminates at system failure

E.g. inability to delivery the rated power or energy

Driven by loss of cells themselves, or loss of access to cells (e.g. because of power converter

failure)

Assumptions and limitations

Failure rate of components is constant in a state

But failure rates can change between states

E.g. as components fail, other components work harder13

Comparing reliability of options 1, 2 and 3

14

Higher relative

electronics

reliability

(or cells age)

Higher cell

reliability

Constant

temperature Including

thermal model

Cell failure rate/base failure rate

Syste

m M

TT

F (

ho

urs

)

Hardware

15

Lithium titanate

20Ah cells (50 Wh)

12 cells in a 3U case (600Wh)

144 cells

in a rack

(7.2kWh)

System architecture

16

Output waveforms

17

Output voltage

Grid current

Distributed balancing

The CHB circuit provides ‘free’ balancing at the cell level as well as DC to AC conversion

Balance cells in a module

Balance modules in a bank

Balance banks in a system

This limits the exchange of information at the cell level

c.f. trying to manage 20,000 cells from a

central controller

Power hardware is ‘flat’ but communication

and management functions are ‘layered’

18

Balancing video

19

Conclusion

Cell-based energy storage systems need a power electronic system of one form or another to exchange power with the grid

Fundamentally to provide bidirectional DC-AC conversion

Perform balancing and deal with cell failure

[typical cell voltage] <<< [grid connection voltage]

Massive series strings are a bad idea

There are more ‘intelligent’ ways to organise a system that give better cell capacity

utilisation and much high reliability

Some power electronics circuits provide ‘direct AC synthesis’

Distributed rather than centralised (or monolithic) converters

There are lots of interesting algorithmic challenges attached to large storage systems

20

Related publications

E. Chatzinikolaou; D. J. Rogers, "A Comparison of Grid-connected Battery Energy Storage System Designs," in IEEE

Transactions on Power Electronics, 2016 (in press). doi: 10.1109/TPEL.2016.2629020

E. Chatzinikolaou and D. J. Rogers, "Cell SoC Balancing Using a Cascaded Full-Bridge Multilevel Converter in Battery

Energy Storage Systems," in IEEE Transactions on Industrial Electronics, vol. 63, no. 9, pp. 5394-5402, Sept. 2016. doi:

10.1109/TIE.2016.2565463

E. Chatzinikolaou and D. J. Rogers, "Electrochemical cell balancing using a full-bridge multilevel converter and pseudo-open

circuit voltage measurements," 8th IET International Conference on Power Electronics, Machines and Drives (PEMD 2016),

Glasgow, 2016, pp. 1-6. doi: 10.1049/cp.2016.0259

C. A. Ooi, D. J. Rogers, and N. Jenkins, “Balancing control for grid-scale battery energy storage system” in Proceedings of

the ICE Energy, vol. 168:2, pp. 145-157, 2015. doi 10.1680/ener.14.00041

Whitepaper: “UK Research Needs In Grid Scale Energy Storage Technologies” (Brandon et al.), http://energysuperstore.org

21