daniel gilbert - epuron solar - how cloudcam has changed the operation of two remote,...
TRANSCRIPT
Outline
1. Epuron & TKLN site background
2. How CloudCAM is working to take on the role of smooth solar integration
3. Tidy hardware and lots of smarts - the physics of clouds
4. Remote sensing is flexible, can be updated in firmware and hardware and sneaks in at a fraction of the cost
5. Results to date at two sites
Wind Power (NSW) Solar Power (NT)
Epuron Pty Ltd
A leading Australian renewable energy company
Cullerin Range Wind Farm
Gullen Range Wind Farm
Silverton Wind Farm
Rye Park Wind Farm
Liverpool Range Wind Farm
Yass Valley Wind Farm
TKLN Solar (1MW)
Uterne Solar (1 + 3.1MW)
Yulara Solar (1.8MW) Total ~7MW Solar PV
TKLN consists of three projects located in Ti Tree, Kalkarindji and Lake Nash, all within the Northern Territory.
Together they sum to approximately 1MW
High penetration PV into diesel mini-grids (80% approx. max instantaneous)
Designed together by Epuron, PWC, CAT Projects and included federal government funding
Each site has a Grid Stability System (GSS) designed by MPower for short term solar smoothing
Each site has a 20 year PPA with PWC which includes ramp rate control
TKLN Site Background
We have been operating for > 3 years
PWC & Epuron relationship remains engaged & collaborative
PV & control hardware in good condition
Upgraded inverter communications to a cluster controller
Lead Acid VRLA batteries performance has degraded
More difficult to meet the ramp rate required in the PPA
Affecting yield
Difficult to manage large battery banks in remote locations
Replacement of batteries is a (slow) process currently underway
Current Status
Epuron sister company Fulcrum 3D Pty Ltd has developed CloudCAM
CloudCAM is an optical camera that tracks clouds in the sky
ARENA 50% funded R&D project as part of the Emerging Renewables Programme
Epuron Solar Pty Ltd is a project partner
Installed easily on to a vertical mounting pole on the side of existing structures
Includes CloudCAM optical fish eye camera, pyranometer, temperature and humidity, logger, remote comms
CloudCAM has now been installed at Ti Tree and Kalkarindji and integrated with the existing solar power station control system
What is CloudCAM
How CloudCAM is working to take on the role of smooth integration
Historically we have used lead acid batteries to smooth the solar generation during cloud events:
Cloud detected; predictive
ramp-down commences; cloud
clears; ramp up and resume
normal operation
Note different ramp-
up and ramp-down
rates are allowed for
in this design.
Cloud detected; predictive
ramp-down occurs until solar
output reaches “safe” level
Cloud clears; ramp up and
resume normal operation
1
2
3
1
2
3
How CloudCAM is working to take on the role of smooth integration
We are now using CloudCAM as an alternative method to deal with cloud events:
CloudCAM at Ti Tree and Kalkarindji are currently working in ‘NowCasting’ mode
CloudCAM takes an image every 6 seconds and runs an algorithm to determine if it is currently cloudy or clear
The CloudCAM hosts a Modbus TCP register which is queried by the TKLN control system
This ‘clearsky’ flag is used by the TKLN control system to ramp up or down solar output at controlled rates
Thumbnails of images are sent every 6 minutes to Fulcrum3D server
CloudCAM will soon be operating in ‘Prediction’ mode
Each cloud is individually identified and mapped
Movements of clouds is tracked between images and predicted
CloudCAM will give an output for the solar resource in 30 sec, 1 min, 5 min etc
Yield lost by reduction of solar output before a cloud is offset by not needing to inefficiently charge and discharge batteries
Using CloudCAM instead of batteries
Tidy hardware & lots of smarts – the physics of clouds
Types of sky & clouds
Examples
Impact on solar output
Difficulty in detection
Speed of clouds as function of height
Height of clouds relating to shadow location
Types of sky & clouds
Clear
Can set solar to maximum power
Wispy clouds
Have minimal impact on solar output
Are difficult to detect
Sometimes purposely not detect due to minimal impact on solar output
Often looks like glare
High clouds
Has small - medium impact on solar
Increases brightness in area around sun
Types of sky & clouds
Partly Cloudy
Full, dark, lower clouds
Distinct cloud groups
Very high impact on solar output
Easier to detect but also can be fast moving
Very dark regions on bottoms of clouds can look similar to ‘clear sky’
Cloudy
Full cloud cover
Medium impact on solar output
Can be minimal unless break in clouds occur
Difficulty in detection
Clouds at different heights can move independently which can make cloud
detection and prediction more difficult.
See example below from 25th Feb 2016 at Ti Tree
Speed of clouds as a function of height
The speed of clouds changes as height increases
This means the expected time before cloud events changes upon cloud height
Geographical location and time of year changes cloud speed
For the NT in Australia, wind speed is faster in winter
This is good because this is the dry season
Data for graphs on the right from the International
Standard Atmosphere ISO 2533:1975
Speed of clouds as a function of height
Transit time of clouds can then be determined as a function of cloud height
The graphs on the right are the time taken for a cloud to travel across ¼ of the camera lens
This kind of information is important for transitioning from ‘NowCasting’ to ‘Prediction’
The vertical lines represent 5 minutes and 10 minutes
Data for graphs on the right from the International
Standard Atmosphere ISO 2533:1975
Tidy hardware & lots of smarts – the physics of clouds
Intelligent part of CloudCAM mostly on the software side
Future development of CloudCAM can be tested historically to previous images
This can be applied to existing installations remotely
Remote sensing is flexible, can be updated in firmware and hardware and sneaks in at a fraction of the cost
CloudCAM can be easily installed, attached to existing buildings, containers or met towers. All it requires a fairly unobstructed view of the sky
CloudCAM comes with the optical camera, the pyranometer and the temperature and humidity sensor which connect to a Fulcrum3D designed and built logger
F3D logger is then connected to the local network via LAN cable and to a 12V power supply
Local hardware is reliable and robust without moving parts
CloudCAM software operates independently & locally with the local PV control system
CloudCAM firmware can be updated remotely
Remote sensing is flexible, can be updated in firmware and hardware and sneaks in at a fraction of the cost
When TKLN was built, the GSS at all three sites cost a combined ~$1M
CloudCAM costs in the order of ~$30k-$40k
CloudCAM can drastically reduce the need for batteries, or remove them altogether depending on the type of project
Perhaps the best combination, depending on the project, is high penetration of PV into diesel mini-grid with much smaller (and affordable) battery support with a CloudCAM
Epuron Solar is currently trialling the CloudCAM with various levels of battery support to see what works best
Results to date at two sites
Yield was decreasing towards the end of 2015 due to battery degradation (more than usual for the wet season)
CloudCAM was first operational in December 2015 but not integrated with the solar control system until the start of January 2015 at both Ti Tree and Kalkarindji
For January 2016 in comparison to December 2015 at Ti Tree, CloudCAM has caused an 4% increase in yield, normalised for variance in monthly irradiation
For January 2016 in comparison to December 2015 at Kalkarindji, CloudCAM has caused an 5% increase in yield, normalised for variance in monthly irradiation
Both sites have had significant decrease in battery cycling which will help preserve battery lifetime
Results to date at two sites
CloudCAM detecting
clouds then clear sky.
Minor yield lost
Complete elimination of
battery use on this day
Results to date at two sites
Significant fluctuations
avoided; minor loss of
production
Dramatic reduction in
battery cycling; dramatic
reduction in battery
capacity required