Solar is a dynamic energy solution trying to make its way in a grid designed for one-way traffic, writes Jonathon Dore of Solar Analytics. Yes, there are a few teething issues – but solutions can be found.

As penetration of rooftop solar PV increases, we are beginning to see the challenges associated with an increasingly distributed and variable source of generation. Across the National Electricity Market (NEM), with rooftop and large-scale solar making up just over 5% of supply, solar variability is still easily smoothed over by other generation sources. At the local level, however, with some areas exceeding 50% of homes powered by solar, we are seeing the impacts of supply excess in distribution networks that are still operating on the needs of high demand.

These challenges manifest primarily as local voltage fluctuations and present the first major limitation on solar deployment and production. What we learn in solving these local issues will be instructive for how we manage supply and demand across the whole market in the future, as solar becomes the dominant source of energy.

Dr Jonathon Dore is head of data analytics and research at Solar Analytics. As a solar specialist for nearly 20 years, Jonathon has previously worked in R&D and manufacturing in Australia and Germany and has undergraduate and postgraduate degrees in photovoltaic engineering from UNSW. 

Voltage and delivery

Voltage is a measure of electrical potential. The higher the voltage, the more current will flow for a given resistance. If we consider resistive appliances such as electric hot water, the voltage will determine how much current flows and thus how much power is dissipated, since power equals voltage times current.

Electricity moves through the grid at different voltages. Over long distances, we use very high voltages, which allows the current to be low. This minimises the resistive losses in the wires, since power lost is equal to current squared times resistance. As the electricity gets closer to the consumers, the grid is broken up into hundreds of thousands of smaller circuits, which serve a collection of consumers, for example a suburban street.

The voltage is stepped down by transformers to lower values, raising the current proportionally so the same power is available. This increases safety but also increase the losses.

Voltage standards define the range in which voltage should be delivered to our homes. In Australia, the target voltage is 230V, with an operating range of +10% and -6%. That means the distribution network service providers (DNSPs) who operate our electricity grids are required to keep the voltage between 253V and 216V. The voltage standard is helpful for appliance manufacturers to design equipment that operates optimally and safely within that range. 

The power required by homes and businesses along the circuit determines the current that needs to flow through the wires. The resistance in the wires causes power loss and we see this as a gradual reduction in voltage along the line, just like a ball rolling along the ground gradually slows down as frictional losses reduce its energy.

For this reason, network engineers have traditionally set the transformers such that the voltage starts at the higher end of the required range, allowing for the gradual reduction so that the house at the end of the circuit still sees a voltage above the bottom end of the range, even in times of peak demand.

PV impacts voltage…

With around a quarter of all houses now fitted with rooftop solar PV, often generating more power than is being consumed, current is increasingly being exported back up the local circuit toward the transformer. With the voltage at the transformer being more or less fixed, this means that to allow for the voltage losses in the direction of current flow, the voltage at the end of the circuit needs to be higher than at the transformer.

The PV inverters effectively push up the voltage at the point of connection to force the current back upstream. But since the voltages at the transformer have already been set high to start with, the voltages at the end of circuits with a lot of exporting PV are often exceeding the upper limits.

…and voltage impacts PV

From a purely technological standpoint, it doesn’t really matter what the grid voltage is. The PV array will generate DC power at a voltage determined by the configuration of the panels – usually between 100 and 600V for residential systems.

The PV inverter converts the DC voltage to an AC voltage similar to that of the grid and converts the DC to AC current such that the overall power is much the same as produced by the panels (minus conversion losses in the inverter). So regardless of whether the grid voltage were 200 or 300V, you could theoretically get about the same output power.

However, Australian standards for PV grid-connected inverters (AS4777) specify how they should respond to high or low grid voltages (and to high or low grid frequency). In the latest version, released in 2015, inverters can respond in a number of ways. If the 10-minute average voltage is above the upper limit, they must either: A, shut-down until the voltage returns to the operating range; B, reduce their power output proportionally to the excess voltage (Volt-Watt response), or; C, absorb reactive power proportionally to the excess voltage (Volt-VAr response).

Options A and B help to reduce the voltage by limiting the current flow towards the transformer, thus reducing the voltage losses along the circuit and hence reducing the voltage that the inverter needs to set to push current upstream.

Option C helps by acting as a pseudo load. When the AC current sine wave is perfectly in sync with the AC voltage sine wave, 100% real power is available. When they are offset, a portion of the available power is “reactive”. Reactive power is needed for things like rotating motors and there is significant reactive power consumption in many industrial applications but very little in most homes.

PV inverters are clever pieces of power electronics and can absorb reactive power such that it creates a voltage loss along the power line to help offset the voltage rise from the exported real power. If the inverter has spare capacity (the power generated by the array is lower than the rated capacity of the inverter), reactive power can be absorbed or generated with negligible impact on PV production. However, if the inverter is at capacity, it has to reduce the PV production to allow for the reactive power.

Similar requirements exist when the grid voltage is below the lower limit, although in that case shutting down the inverter can exacerbate the problem and this is currently under review.

Inverters installed before 2015 mostly operate under the previous version of the standard, which just specifies response option A. There are also provisions for very high and very low voltages, in which the inverter needs to shut down within 0.2 seconds due to the likelihood of a major grid disturbance. 

The DNSP responsible for the grid in which the inverter is connected specifies the values of high and low voltages (as well as the very high and very low voltages) at which these responses kick in. These are generally quite close to the required operating range of 216 to 253V.

Voltage monitoring

Until recently, there has been very little visibility of voltage values at the point of connection. DNSPs generally have monitoring at the sub-station transformers, which are further upstream from the local transformers. So this means they can monitor the voltages in, for example, a given suburb to ensure there are no major faults. Very few local transformers have monitoring, so to check the voltage there or at a house or business, a technician usually needs to physically measure it on site. This makes it difficult to ensure that voltages are kept in the operating range. 

The increased roll out of smart meters adds the potential for greater visibility, but getting access to these data can be difficult and expensive. 

Thankfully, voltage is measured as part of the solar monitoring and energy management system that Solar Analytics is providing to a growing number of customers around the country. The system typically uses energy monitors made by Wattwatchers, which measures voltage, current, frequency, real and reactive power and energy from each phase over periods down to five seconds and sends the data to Solar Analytics cloud servers.

Figure 1: Proportion of sites measuring voltage above the upper limit at least once on fewer than the given number of days, sampled from 1,000 sites spread around Australia over all of 2018.

As this energy database grows, we can start to see some interesting trends.

Figure 2: Proportion of sites measuring voltage below the lower limit at least once on fewer than the given number of days, sampled from 1,000 sites spread around Australia over all of 2018.

Figure 1 shows the proportion of sites on the Solar Analytics fleet at which voltage was measured above 253V at least once on each of a given number of days in 2018. Around 85% of sites experienced over-voltage at least once in that year. About 50% of sites measured over-voltage on more than 50 days and more than 25% of sites had over-voltage on more than half of the days in the year.

At the other end of the scale, Figure 2 shows the same statistics but for measurements of voltage below 216 – the lower bound of the range. This shows the number of days out of spec to be much lower. For example, less than 2.5% of sites had more than 50 days where voltage was under the limit. 

So we see many voltage excursions at the upper end but very few at the lower end. This makes sense given the earlier explanation that transformers have traditionally been set high to allow for voltage loss during peak demand and shows that they have not adjusted to the new (and increasing) reality of reverse current flows. 

Naomi Stringer, a researcher at the Centre for Energy and Environmental Markets at UNSW, studied voltages hour by hour using anonymised data shared by Solar Analytics as part of a research collaboration supported by the federal government’s CRC program.

Figure 3 shows the data for January 2017 for Queensland. While there is some variation throughout the hours of the day, the overwhelming majority of voltage measurements are well above the target 230V and are most often between 240-250V, both within solar hours and at night-time. Similar trends were found for NSW and South Australia.

Action on over-voltage

Figure 3: Box and whisker plot showing median, interquartile range and outliers of voltages recorded in Queensland in each hour during January 2017. Source: Data driven exploration of voltage conditions in the low voltage network for sites with distributed solar PV, 2017, Stringer, Bruce and MacGill, Asia Pacific Solar Research Conference paper.

It is very difficult to know before purchasing a solar system whether the voltage at your location is high and installers should do an on-site inspection before designing and quoting a system, including a voltage measurement. 

Active monitoring software is the best addition to a system for owners to understand whether their investment is working properly. A system like Solar Analytics can detect loss of production due to over-voltage and alert the owner. The data shown above suggests that many of the two million rooftop PV systems in Australia experience some production loss due to over-voltage without anyone knowing. 

Your first point of contact should be the company or person who sold you your system. If you have monitoring, they should be able to access the data too and help confirm what’s happening. They can check the inverter settings to make sure the right response modes are active and that the voltage thresholds are correct. If the problem persists, contact the DNSP.

I had first-hand experience with this for my own system and had a same-day response from Endeavour Energy, who sent a technician out to adjust the local transformer to a lower voltage (a process known as a tap change).

Figure 4 shows the voltages measured at different times of the day throughout November 2018, before and after the tap change, which conveniently occurred around the middle of the month. From that point on my system began performing much closer to the expectations calculated by my Solar Analytics monitoring.  

Figure 4: Voltage measured at a meter board of a rooftop PV system (above) and energy produced vs expected from Solar Analytics Performance Dashboard (below) before and after adjustment of the local transformer turns ratio.

Networks’ dilemma

DNSPs are large, complex, highly regulated organisations with a massively important job to do. Depending on who you talk to, they are either well aware of present extent of this problem or very surprised by the data shown above. Likewise, they are either very keen to manage it in a way that helps to accelerate the transition to a clean energy future or believe that solar alone is to blame for voltage problems and the way to deal with it is to stop approving connections. 

It is understandable that voltages are set cautiously high since the PV inverter standards put a brake on the system at the upper end, but no such brake exists at the lower end. On extremely hot days, when people are coming home and turning on air-conditioners at the same time, local demand peaks and voltage goes down. No standard exists to turn off air-conditioners or other loads when voltage dips below the threshold, although consumers can be incentivised to do this.

Several demand response trials have been conducted or are underway that involve incentives for customers either manually or automatically turning off air-conditioners or other devices for brief periods during peak demand times. These have all shown potential for demand reductions, which will help voltage management at the local level and help meet broader supply at the market level. Consumer engagement is difficult, however, and Solar Analytics is leading a project with Energy Queensland, SAPN and Wattwatchers to tackle demand response from a consumer perspective. 

Looking for a solution

Even if, through demand response, voltage can be managed just as well at the low end as at the high end, it will still be a difficult exercise to shift voltages to the middle of the range. Manually adjusting thousands of local transformers one customer-request at a time is clearly not an efficient exercise and ultimately increases costs for everyone. A more coordinated response is required.

South Australia, with the highest rooftop PV penetration in the world, is experiencing many of these issues first. Consequently, over-voltage enquiries to SAPN have been increasing, as shown in Figure 5. While SAPN imposes a 5kW per phase export limit of PV systems, its LV management business case shows that this is not a long-term sufficient limitation and proposes a dynamic export limit informed by increased transformer monitoring and live system data. Such data can come from smart meters, from comms-enabled PV and battery inverters and from monitoring systems such as Solar Analytics. 

It’s expected that all DNSPs will rely on vastly increased live data visibility to aid their future planning as well as manage operations, enabling them to target areas in need of upgrades or voltage adjustments in a coordinated and more cost-efficient manner. Figure 6 shows Solar Analytics sites in the ACT on the ActewAGL network, when 100% of sites simultaneously experienced a step in voltage of more than 3 Volts one day in March, 2018, indicating some sort of grid disturbance. Live streams of such data will help network operators respond faster to problems, benefiting consumers. 

Solar Analytics is providing some of this anonymised data to the Australian Energy Market Operator through an ARENA-funded research project in order to investigate how PV systems respond to grid disturbances, including voltage spikes or dips. This has helped to inform the recommendations in AEMO’s Technical Integration of Distributed Energy Resources Report. For example: “There is benefit in expanding voltage disturbance withstand capabilities … of DER as much as possible, taking into account safety considerations…”

What does the future hold?

Wattwatchers CEO Gavin Dietz sees a future for consumer-side-of-the-meter energy technologies where household and business data will regularly be shared with the industry – electricity networks, retailers, aggregators and other service providers – on commercial terms as part of normal operations.

“Within a decade there’ll be dramatically more energy data available through the cloud, and the utility billing data that dominates today will only be a fraction of the total,” Dietz has said. “The bulk of it will come from consumer-owned assets, including inverters, EV chargers and smart devices … The flow of this data via APIs will be critical to the trade in electrons in increasingly distributed electricity grids.”

Rooftop solar is continuing to grow at rapid rates, and about 250,000 systems will be added this year. This is certain to continue, primarily because it is the cheapest form of electricity. A recent study by the ISF, CEEM and APVI found that rooftop potential solar capacity is over 20 times what we have today, so space is not a limitation. This presents an important role for regulators to ensure that the benefits of cheaper energy can be shared by all consumers and by the various participants in the energy system. 

Batteries and aggregators

As battery storage falls in cost we will see increased deployment of storage, with AEMO predicting around 3GW cumulative battery capacity to 2030. To an extent, this can alleviate voltage issues by smoothing out the generation and demand profiles, particularly with the highly synchronised evening demand.

During the day, an economically optimal sized home battery is unlikely to take up all of the excess solar energy, so the combined peak export later in the day will likely be unchanged. This is where smart control algorithms can help. Such algorithms could learn to forecast high voltages and other local constraints and delay charging until those times, when it is most valuable.

This relies on sufficient incentives, for example avoidance of production curtailment and export limits or payment for grid support services. Likewise, providing incentives for batteries to alleviate network constraints in both directions could be more cost-effective than network augmentation. 

It is likely that such control will be provided by third-party aggregators, who manage the communication and value exchange between inverters and networks or retailers and can coordinate large numbers of systems to optimise the outcomes for all. Many more start-ups like Solar Analytics or Reposit Power will compete, with those who understand how to engage the end user likely to win out. Platforms such as Greensync’s Distributed Energy Exchange (deX) will provide a marketplace for this value exchange, enabling the smallest consumers right through to the largest retailers to benefit. 

Virtual power plants and EVs

An example of such coordination is a virtual power plant (VPP), where many batteries can supply a large amount of power when required. This is particularly valuable when demand spikes and high gains can be made on the spot market or in contract with retailers. However, if the high demand is not replicated in the areas where the batteries are located, then localised problems may occur due to the large simultaneous export. This risk is one of the motivations behind the dynamic export limits proposed by SAPN.

Coordination will become even more important as electric vehicles become more prevalent. This is one area in which Australia is lagging and forecast take-up is almost impossible given the heavy reliance on coordinated charging infrastructure and the likely need of subsidies to kick off the local market. Delays now could make the eventual take-up even more rapid once prices or availability make it inevitable.

This rapid change will create an even greater challenge for electricity systems to keep up. The synchronised demand of cars arriving home and being plugged in for charging will add to the existing peak unless incentives or regulations are provided to delay this load to a low demand period and coordinated control systems are created to react optimally. 

Peer-to-peer power

A further piece of the puzzle is local sharing of electricity. Presently, network costs don’t take into account how far the electricity travels, so buying electricity from next door costs the same as buying it from a generator a thousand kilometres away. Local supply and demand constraints could be alleviated by taking proximity into account, creating a natural incentive for consumers to shift loads to when their neighbours are producing energy, or conversely for battery systems to export when there is higher local demand.

Again, this is unlikely to be achieved by human decision making and requires automated control systems to act according to user preferences. Technology companies like Power Ledger and Enosi are creating blockchain algorithms to manage such trading, while trials engaging consumers are being conducted by AGL, Solar Analytics and others.

So the future will see many solutions to the emerging constraints of the clean energy transition but also many new challenges which keep shifting the goalposts for everyone involved. What is clear is that high-quality, high-volume data, connectivity and coordination will play an increasingly important role and that at the centre of all this is a person who just wants to be comfortable and connected at a reasonable price.

Dr Jonathon Dore is head of data analytics and research at Solar Analytics. As a solar specialist for nearly 20 years, Jonathon has previously worked in R&D and manufacturing in Australia and Germany and has undergraduate and postgraduate degrees in photovoltaic engineering from UNSW.