Skip to content

Rooftops and satellites: Predicting the solar future

Last year, a “record” was set when the overnight electricity demand on Western Power’s network was greater than the following day’s maximum demand.

Day has become night. Or night has become day. 

It’s one of the dramatic changes Western Power is seeing from increasing uptake of solar panels.

While welcome, this increasing solar penetration presents challenges for the whole community – and the network.

They say every cloud has a silver lining, but what if clouds suddenly cut energy production from rooftops? A sharp drop in electricity supply can affect everything from appliances, to power in neighbouring homes without solar panels to other generation on the network.

While we wait for battery technology to become viable at a grid level to provide back-up for these fleeting sharp changes in electricity supply, Western Power has linked with a satellite monitoring company and researchers at Australian National University (ANU) in a world-leading bid to forecast solar energy supply.

Taking satellite cloud-tracking technology from emerging tech company Solcast, researchers at ANU will combine advanced modelling techniques with Western Power’s network-wide data to forecast solar power output up to four hours in advance.

It’s a creative solution to ensure household power supplies remain safe and reliable, while embracing the opportunities. 

In this extended post, Dr Nick Engerer of the Australian National University explains how our teams are bringing together satellite technology, high volumes of data and forecasting models to ensure we can enable a strong solar-powered future:  

By Dr Nick Engerer
Australian National University (Fenner School) 

On 14 August, 2016, Western Power set a new solar record. With 350-400MW of solar power being produced by its customers, its daily minimum electricity demand was lower than the minimum demand the night prior. To clarify, what this means is that this was the first time daytime energy use was lower than the night time usage.

While this ‘record breaking event’ may sound quite ordinary to the casual observer, it is actually a very clear example of why quantifying solar power generation across Australian electricity networks is becoming a priority.

Solar power contributions to our energy mix were relatively insignificant only five years ago - now they are a substantial contributor to the electricity supply-demand balance.  So large, in fact, they can make the daytime demand look like a night-time load profile during a sunny day.

Even more challenging are the fast-moving, fast changing cloud cover conditions caused by cold fronts moving in from the southern Indian Ocean to the west. With an impressive one in five Australian homes having a solar photovoltaic (PV) system installed, there is now a need for technologies that can quantify the total power output from these systems and predict any big changes in that energy generation well ahead of time.

Western Power teams up with the ANU and Solcast

Western Power is taking action to confront the challenges of integrating these large numbers of intermittent electricity generation sources by teaming up with my research team at The Australian National University (Fenner School). Through funding from the Australian Renewable Energy Agency, we are collaboratively building applied solar forecasting technologies which will remove the uncertainty in power generation from these solar installations.

The key source of uncertainty with these solar systems is that the majority of them are not actively monitored. This means that 218,000 electricity generators operate in their network, but Western Power doesn’t know their past, current or future contributions to the electricity supply. This creates a key limitation in the management of their power output - because solar energy generation varies in intensity throughout the day, is not available at night, and is routinely interrupted by cloud cover. Our project will confront this lack of information head-on, by modelling total solar energy generation through the use of two sources of information.

Combining satellites and live solar PV data

First, through partnership with our start-up company Solcast, we have applied state-of-the-art solar radiation modelling techniques to the brand new Himawari 8 satellite. Himawari 8 is a geostationary weather satellite that tracks cloud cover over Australia at 1km^2 resolution and updates every 10 minutes. This state-of-the-art piece of equipment became operational in 2016 and is the first satellite with such high resolution data.

Solcast has built forecasting capabilities using this satellite, which can forecast the available solar radiation anywhere in Australia, up to four hours in advance, by tracking and predicting cloud opacity and motion. Though the use of an ensemble (a group) of weather models, it then extends these forecasts out to seven days ahead.  

My R&D team at the ANU then apply Solcast’s operational solar forecasting data to produce PV system power output forecasts through partnership with Western Power.  

Western Power’s engineering team provides us with detailed installation data for these small-scale solar systems, which we use to build a virtual model of their PV fleet.  

Starting in the coming weeks July 2017, we’ll be delivering these forecasts to Western Power’s control rooms operationally, with continual refinement and improvement of the service through the end of the project in 2019.

While solar forecasting for energy networks is not entirely new, the level of network level visibility we’ll provide Western Power is unprecedented.  Their engineers will be able to zoom all the way through their network, from zone substation, to feeder line, to each distribution transformer and see the current, past and future solar power generation by that network asset.

An example of our satellite-enabled solar PV nowcasting capability is shown in the above demo built by ANU Engineering R&D student Will Barker.  Satellite-based radiation estimates are used to generate predictions of PV system power output using data at distribution transformer level, before aggregation back up to postcode regions.  At far left is imagery from the Himawari 8 satellite; at right, the total aggregate power output across all of the solar PV systems in the Western Power dataset.

One of the key ways that we will provide such high levels of detail with confidence is through the use of real-time data feeds from tens of thousands of solar PV systems via partnership with solar inverter companies SMA Australia and Fronius.  

These data points will allow us to validate our forecasting methods and make improvements to capture system-level effects such as shading or soiling (aka dirty solar panels!).  

High resolution solar forecasting as an enabling technology

Looking forward, having better solar PV intelligence across their network, will enable Western Power to get pretty creative in its solutions for renewable intermittency and electrical grid infrastructure costs. With high resolution, high accuracy solar forecasting integrated with their operations, variability in solar can now be matched with the charging/discharging batteries and the shedding of load via demand management initiatives.  

We can, for example, group solar forecasts by zone substations (see next graphic), in order to predict the total changes in the load on the network that are caused by cloud cover - this could then be compensated for by dispatched power from a battery system sited nearby.

Same as graphic, 1, except only the solar PV sites connected to the Geraldton zone substation are included.  These types of network-centric forecasting approaches are the key enabling information for coordinating solar variability with smart grid solutions such as battery storage or demand side management.

At Western Power, the impetus for finding such creative solutions comes from the top. In a 2016 interview, Western Power’s CEO Guy Chalkley acknowledged there were risks posed by emerging technologies such as solar panels and batteries, but also stated that he believed there is opportunity to combine solar panels and batteries to provide a cheaper and more reliable supply for some customers. 

In response to this leadership, Western Power has already begun some of Australia’s most innovative solar and storage projects, including stand-alone power systems, a renewable energy microgrid and a network battery.    


With a progressive future on the agenda, the ANU and Solcast are moving quickly to support Western Power with high resolution solar forecasting data. By the middle of this year, Western Power will have access to a trial service for use in their control rooms and network planning.  

Through their feedback, this service will undergo continual improvement over the course of 2017 through 2019, ensuring that this technology is integrated directly with their decision makers in a meaningful way.

It’s all part of a bigger picture plan to enable our distribution networks for the solar powered future ahead, ensuring that more solar technologies can continue to be added to the electricity grid.

I will be sure to update you with our progress; we are excited to be working with such a forward thinking partner in the area of operational solar forecasting.


Dr Nick Engerer is the Chief Investigator on a $3.6M ARENA distributed solar modelling project at The Australian National University (Fenner School). He is an expert in the field of solar radiation and distributed solar PV modelling whose personal objective is raising the maximum penetration levels of solar in our electricity networks. You can connect with Nick on LinkedIn or follow him on Twitter.

Share this article