This document summarizes the use of lithium-ion battery energy storage systems in Queensland's regional electricity distribution network. It discusses how a system called GUSS (Grid Utility Support System) was developed to provide energy storage and help regulate voltages on long rural feeder lines. Near real-time monitoring of the batteries provides insights into their performance and how they impact the network. Sensor data from the batteries and surrounding infrastructure is analyzed to identify issues and optimize the systems. The data collected also helps engineers understand how the batteries age over time in different usage conditions.
(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7
Analytics for utility-operated lithium-ion energy storage
1. Analytics for utility-operated
lithium-ion energy storage
Monitoring the performance of grid-connected
batteries in the Queensland distribution network
Dr David Ingram 7 December 2017
2. Overview
● Regional electricity supply in Queensland and the need for upgrades
● Lithium ion battery terminology
● The Grid Utility Support System (GUSS)
● Remote data collection and sensor fusion
● Examples of how access to near-real-time data solved engineering problems
● Future opportunities
● Conclusions
2
3. Acknowledgments
● Ergon Energy for permission to use
photographs and graphs
● Former colleagues/managers
Steve Richardson
Michelle Taylor
Jason Hall
● Data sources
Spatial data from qldspatial.qld.gov.au
Ergon Energy for sub-transmission and
distribution substation and feeder data
Geoscience Australia for transmission line data
3
Michelle, Jason and
Steve at the 2016
Australian Engineering
Excellence Awards
5. Regional supply in Queensland
● Ergon
728 000 customers
118 000 km of distribution
1 836 500 km2
5.2 customers/km
● Single Wire Earth Return (SWER)
25 000 customers
62 500 km of lines
500 400 km2
0.4 customer/km
5
6. SWER networks
● Characteristics
12.7 kV and 19.1 kV
Long spans with high tension
Highly resistive, R/X is 5 to 10
Long distances
Point-of-load transformers
● Issues
Increased load
Sensitive electronic devices
Voltage regulation
6
Photo by Bernard S. Jansen (Wikimedia Commons),
CC BY 2.5
7. Example of multiple SWERs on one feeder
● “Alpha” Feeder
1890 circuit km
29 000 km2
~900 connections
2.3 MVA peak
● 8× SWER Systems
7
Satellite background from ESRI Global Imagery
10. Introducing GUSS
● 25 kVA inverter
● 4 quadrant (±P, ±Q) operation
● 100 kWh of energy storage
● Skid-mounted
● Remote control and alarming
● On-board data collection
● INMARSAT/3G modem
● Now in Overhead Construction
Manual
10
Inverter Controls
Battery
Photo courtesy of Ergon
Energy
11. Commissioning
philosophy
● New approach
● Used Maryborough workshops
● Remote access tools used
● Proved communication & data
processing systems worked
● Minimised on-site time
11
Photo courtesy of Ergon
Energy
14. First
installation
● “Weir” on a Charters
Towers feeder
● 50 kVA (2× GUSS)
● 80% along the feeder
● Satellite modems
14
Satellite background from ESRI Global Imagery
GUSS
17. Building a battery system
17
StringModuleCell
System
Photos courtesy
of Ergon Energy
18. The need for balance
● Safety critical function
● Cell voltages must be similar
● Worst-case limits battery capacity
● Automatic balancing through discharging
● Intelligent monitoring
SMU monitors all cells in module
BMS monitors all modules
● Series/parallel combo of 784 cells
● Total voltage of 706 V
18
22. Reasons for monitoring
● Monitor string balance
● Provide advanced warning
● Collect engineering data in
parallel with operating data
● Fine tuning of advanced
control algorithms
● Assess effect on network
● Verify that contract technical
specification is met
22
Photo courtesy of Ergon
Energy
23. Technical constraints
● Satellite data is expensive
● Need to be selective
● Couldn’t alter BMS or PCS
software
Safety critical systems
● Pre-process of data before
transmission to reduce size
● Selection of file transfer protocols
Effect of latency
Compression
Cyber-security
23
Inmarsat BGAN M2M
coverage
Photo from WideEye
24. Sensor fusion
● Multiple intelligent devices on board
Energy meter
Power Conversion System
Battery Monitoring System
● Grid monitoring
Isolation transformer monitors
Series step regulators
Automatic circuit reclosers
Power quality meters
● Combine data for greater insights
24
Photos courtesy of Ergon
Energy
Cooper
Industries
25. Data sources and transfer method
● Report By Exception (RBE)
Alarms and events
PCS analogues beyond thresholds
DNP 3.0
● Daily batch transfer by file
String data
Module data
Cell data
SCP
● Daily batch transfer by Metering
Energy meter data
Proprietary metering protocol
25
26. Workflow
● Daily download of data in shared DB
Download each GUSS sequentially
Feed data into corporate repository
● Process the data and generate graphs
Series of charts with different focus
Provide a summary website
Reduces server workload
● Weekly and monthly processing
Examine long term trends
Data archiving
26
28. Battery monitoring
● System-level monitoring
DC bus voltage
State of Charge
Battery string currents
Range of cell voltages
Maximum/minimum cell temps
● Module-level monitoring
Voltage/temp variation from the
median
Mean/median voltages &
temperatures
● Cell-level monitoring
Voltage variation from the median
28
Photo courtesy of Ergon
Energy
29. Performance assessment
● Look at the “fleet” individually and as a whole
Watch for abnormalities
● Too much data to look at in numerical form
Visualisation is key
Catch potential faults early
● Use data from multiple sources
No one source has all the required information
● Visualise data in a variety of ways
29
37. Many ways
to visualise
● Potential module
fault in String 2
● Module 20
● Cell 9
● Outliers can be
masked by large
variation
● Expect this with
daily cycling
37
40. Modelling of battery aging
● Building a library of operation
over a range of conditions
● No two sites are the same
● Cell-level and module-level
logging creates a deep dataset
● Applicable to three phase and
isolated use too
40
41. Optimising control system parameters
● Power system models give initial estimate
● Designed for autonomous and directed operation
● Control systems do interact
● Observed response can be used to tune power models
● Use power electronics to reduce wear on switched equipment
41
42. Data-constrained environments
● “Design for data”
Make key items available
Open protocols
Enable automation
● Utility applications
Automatic circuit reclosers
Step voltage regulators
Energy & power quality meters
● End-user applications
Irrigation
Load control
● Satellite
Medium bandwidth
Very high latency
Expensive
● Internet of Things
Low bandwidth
Medium to high latency
Power is limited
42
43. For further detail
Design and commissioning detail in:
● Transmission and Distribution World
● June 2017, pages 42-48
● www.tdworld.com/distribution/storage-has-vital-role
43
44. Conclusions
● Power electronics can solve capacity and voltage problems
● Use of communications reduces risk of deploying technology
● Remote data collections gives great insight into behaviour of network
● Consider all the constraints of selected communications network
● Better situational awareness improves decision making
● Early indication of problems gives opportunity to plan for repair
● Incorporate performance/condition monitoring at the design stage
44
The grid is built in layers: Transmission (Powerlink), Sub-transmission, Distribution, SWER
Not many customers. Energex has 1.2M in the SEQ corner
SWER network is extremely remote
Voltages based on phase to phase voltages of 22kV and 33kV.
Resistive conductors, with 3 and 7 strand steel common.
19.1kV experiences Ferranti effect, increasing voltage at time of light load.
Shunt reactors used to limit rise, but not switched.
Step regulators are used too. Subject to mechanical wear and tear.
Difficulties regulating voltage.
Modern electronic devices are less tolerant of voltage swings.
“Alpha” feeder from Barcaldine substation
Equivalent to Belgium or Armenia in supply area.
Significant voltage control problems.
Expensive to install bigger conductors and limit to # of regulators that can be fitted.
Black is the 3-phase backbone, SWER branch off from sides. Yellow diamond is SWER isolation transformer.
Technical specification for tender developed by Ergon
S&C Electric integrated inverter/PCS and battery.
Ergon integrated metering and communications.
Joint commissioning and installation.
Transported fully assembled on flat bed truck, can take up to two at once.
Communication over both 3G and geostationary satellite due to remoteness
All units were commissioned and had meters and modems fitted in Maryborough by Ergon.
Data collection and analytics were refined and process verified as working before transport to site
Deliberate decision made to use 3G modem for commissioning rather than hard-wired cable.
Developed work practices suitable for long term support from the office.
Review of all work practices, including scheduling and training was performed.
Take the opportunity to identify all new ideas, and past pain-points.
Talk to everyone that will be involved.
Units roughly 50/50 split between dual and single sites
All over Queensland, in every region
WB and FNQ were used for workshop and lab work, SW, CA, MK, NQ for deployment
10x Satellite, 7x 3G modems. Red is 3G, Blue is Satellite.
Twin units have a ring around.
First GUSS installed on a complicated SWER with many regulators, reclosers and covering a large area.
Modelling indicated that optimum location 80% along one spur
Need to avoid instability of control loops having two units at the same connection point
Collected data from day 0 (commissioning).
Kinnoul GUSS, NE of Roma
New poles installed. 3G communications via Yagi antenna.
Self contained and exposed to the elements.
A good site for experimentation due to fast and cheap communications.
Pays to have somewhere to prototype things.
Terminology that is relevant to upcoming data.
14 cells in module, as 2 parallel strings of 7 cells in series
28 modules per string, 2 strings in parallel
String or system module monitoring is straight forward.
Module level monitoring takes a bit of work
Cell level monitoring is very rarely done due to number of cells in a large system.
But it all comes back to cell voltages and how “balanced” they are.
Many different chemistries
NCA and NMC have more of a slope, so easier to estimate SOC
Cyclic life improved if cells not fully discharged.
Aim for 20% or 30% minimum SoC.
Need to involve people familiar with the product/process/business need.
Communications experts provide secure, reliable and cost effective bearer and IT systems
Analytics people use the data to explore behaviour, report on observations.
Feedback to product team to change data collection or offer insight.
Many data collection systems supplied by vendors are designed for high speed Ethernet.
Design does not factor in cost, bandwidth or latency.
Cannot alter safety-critical software
Fortunately there was an embedded PC on-board that I could use to pre-process data and to act as a secure server.
Good idea to have some sort of general purpose computer, even if it’s an industrial-rated Beaglebone Black.
Time synchronisation is important.
Blend multiple devices, protocols, data stores together
Thermal performance was of great interest
An energy meter was used as an independent check of performance
Tie in observations from reclosers and regulators along the feeder.
The GUSS was installed to meet a system need, so need to assess whether that is being met
Use for model tuning, and look for reduced operation of electro-mechanical devices.
Protocol choice matters.
SCP much faster than SFTP. Does not require ACK of each packet, so not affected by high latency links.
XML might be popular for internet connected products, but CSV and serial rules the industrial world.
Formats that compress well are good.
No human intervention needed to retrieve data
GUSS is an operational system, so access to engineering tools and communication links was tightly controlled.
Common set of graphs produced each day and put on a web server for anyone in company to access.
“Self-serve” page allowed custom time ranges to be plotted
Meets the Operational Technology security architecture.
Module and cell monitoring tracks variation from the group.
Voltage and temperature is expected to cycle during the day due to climate and energy use.
Want to identify outliers or unexplained behaviour
All units are built the same, so are expected to behave similarly.
Any variation should be examined and cause determined.
Bring data together.
Until the operating model is developed human analysis through visualisation is important.
Have time to explore the data.
Visualise using different approaches. Try different software if available.
Mean of 28 modules gives smooth output, but ignore the outliers.
Median was selected as variation reference point.
Dots represent values that are beyond 2x interquartile range; the outliers.
Right-hand bank has less temperature variation.
Box plots laid out to reflect physical configuration of string.
Modules 18, 19, and 28 are cooler than the median.
Only observed in the workshops.
Suspected to be the cold concrete floor helping get heat away from the bottom modules.
Only becomes apparent due to physical representation.
Module 13 is approx. 35mV lower than median.
Both strings behave same overall.
Balancing should bring all other modules down to M13 level.
Can’t change one by itself, but can discharge individually.
Monitor over a week to see if problem persists. Most go away.
19 October to 3 November.
Line voltage with and without GUSS.
Loads not particularly high at that time of year, so inverter not working near max capability.
Don’t regulate to zero deviation, but enough to make an improvement.
Ergon developed software to poll for cell voltages of every single cell, at two minute intervals.
String 2 Module 19 has low voltage on a number of cells. Only 20mV, but something to watch for.
Graph gave enough time to make arrangements for spares to be organised.
Module replaced at next visit. Avoided unplanned outage sometime in the future.
Spotting trends in time series data can be difficult
Distribution analysis, such as histograms or box plots, can give more insight.
Need to understand what the expected behaviour is, and develop analytics to suit.
Intended GUSS behaviour was to go between 30% and 100% state of charge in a day.
Cells vary from 3.3V to 4V over 20% to 100% SOC.
Can see 30 mV variation with 80 mV range, but would be difficult with 700 mV reange.
Very clear with box plot.
Longer term trends looked at as cross-plot of parameters or distributions.
Capacity in kWh vs SOC looks at battery aging
Limits are constraints on charging or discharging put onto the inverter by the battery management system
Also look at network impacts and how hard battery is working
Thermal constraints are significant. Inverter and batteries have different limits & thermal inertia.
No documented models of calendar and cycling aging in an uncontrolled environment.
Collecting as much data as possible now, modelling is a future project.
Locations chosen through power system modelling, particularly V/P and V/Q sensitivity.
Validate this through observations.
Controller parameters need tuning, particular for load sharing of twin units.
Future analysis to optimise GUSS parameters with step regulator control loops and directed load support.
Develop guidelines for data collection and condition monitoring for constrained systems
Design needs to support automated extraction and analysis. Too many products rely on GUIs.
Would be good to have standards-based way of dealing with constrained environments.
Applicable across the energy sector, for customers and utilities.
IoT similar to satellite, but with power a constraint instead of cost.
Still face bandwidth and latency limitations.
For more details on design, selection and commission please have a read of the Transmission & Distribution World article published in June.