6.1 Potential electricity network impacts

An electric power system consists of a number of generators and a number of consumers, or loads. Conventional electricity networks do not incorporate storage (except at sub-cycle time scales). This means the energy generated must be exactly balanced by the energy consumed at all times. For example, if a person switches on an air conditioner (increasing the load), then a generator has to increase its output immediately to compensate. All generators have limits to the speed at which they can respond to a change in load (also known as ramp rate). When generators cannot respond fast enough, the power imbalance causes symptoms such as frequency and voltage fluctuations. Two characteristics of consumer loads help to prevent serious problems for the most part. First, their large scale effects are predictable. Engineers and planners working for utility companies and energy market operators know when to expect large increases in demand, and can schedule extra generation or alter maintenance schedules to ensure enough power is available. Also, these large scale effects are relatively slow moving, allowing time for generation to adapt to demand. Second, the small scale effects of demand, such as the precise time of a customer turning on their air conditioner, happen in a largely independent and uncorrelated manner (Mr Smith does not turn on his air conditioner at exactly the same time as Ms Jones). Consequently, these small scale effects tend to cancel out due to their randomness.

Power utilities have dealt with these effects for many years, and are experienced and capable in maintaining this delicate balance. Just how difficult this is can be seen by the catastrophic power blackouts that occur from time to time, sometimes started by the most innocuous seed event. It may be thought that PV generation produces variation just like loads changing on the network. However, the intermittency of power from PV generation is different in several ways.

First, large scale variation in PV generation is in the form of the classic sinusoidal function rising from a minimum in the morning to a maximum at noon, then tailing off again to sunset. This is in contrast to residential loads, which tend to peak in the evening. The whole system load, however, better correlates to the classic sinusoidal pattern associated with solar generation. Second, the small scale variation is caused by clouds and is not easily predictable. Although research efforts are being made in this area, it is still not clear what weather monitoring equipment is needed to do this accurately and efficiently, neither is it known which prediction methods are the most suitable. Third, power utilities do not have years of experience in dealing with PV generation and are therefore wary of adopting new technology that could upset the delicate balance that is the daily reality for maintaining a reliable power supply.

Intermittent power generation can have a variety of effects upon the electricity networks to which they are connected. These effects are strongly influenced by the type of network and the amount of intermittent generation, but mostly fall into the categories of stability and voltage effects. High PV system penetration can be characterised by a single large generator in a given network area, or a large number of small generators in a given network area when compared with the load in the same network area. In addition to technical phenomena, there are effects upon the energy market of trading power generated by solar resources.

The successful integration of high penetration solar power into the Australian electricity network is far from assured. The recent rapid uptake of rooftop solar panels around the country has raised concerns amongst various utilities on the impacts of high penetration solar PV. The installation of additional renewable generation has been stopped in certain areas in Queensland and Western Australia. In Queensland, some new applications for rooftop solar systems have been rejected and in Western Australia, limits have been set in some areas on how much intermittent renewable energy can be installed in a system without affecting the power supply. Horizon Power in WA has been rejecting applications for new installations in Exmouth and Carnarvon, and restricting new installations in Broome and Leonora [84]. The issue here is not specifically of intermittency but of voltage rise up to the maximum allowable level triggering the ‘tripping out’ of the individual systems. This is a conservative response to a lack of information on the potential network problems intermittent renewable generation could cause or what mitigation measures are available to accommodate.

6.1.1 Voltage effects

Voltage effects are illustrated by the figures in Section 9.2.2. However, voltage variations are generally local effects that can cause inconvenience to a small number of people. Momentary high voltage can cause damage or shorten the life of equipment, and this is generally regarded as a symptom of adverse network issues. Voltage variations at the level of four volts, as illustrated in section 9.2.2, are unlikely to cause any noticeable network issues. However, utilities routinely maintain voltage near the maximum limit to avoid dropping too low when supplying large loads. When solar PV installations feed power back into the network, a small increase of four volts can push the voltage above this maximum limit. Although voltage variations are already caused by load changes, and utility companies already have mechanisms in place to deal with this, adding local generation can increase the range of voltage variation. The voltage limits are fairly broad and can be managed, but utility companies need to change voltage limits and practices in order to deal with this. At present, utility companies do not routinely monitor voltage at customers’ premises, so do not always possess the information required to accurately set local voltage levels. This is set to change with the installation of smart meters at residential properties. Amongst other things, these are capable of monitoring voltage. In this case, utilities may come under increasing pressure to rectify voltage issues. This may result in utilities exerting tighter control over voltage levels because of their regulatory requirement to comply with high and low voltage limits for the network under quality of service provisions.

PV installations have another problem related to voltage, because each inverter has a voltage operating range. If the voltage at a customer premises is already near the maximum tolerated by the inverters of the PV systems, a small increase in voltage can exceed the maximum for the inverter, which causes the inverter to shut down. This will lead to the voltage falling again, and after some time the inverters may reconnect. So even a steady insolation level can result in fluctuating voltage and a barely functioning PV system. Many customers would be unaware this is happening as they have no technical knowledge in the area, and utilities may regard it as the customers’ problem. This issue is primarily due to inverter control settings, not inverter limits, which are in place to satisfy grid/utility/standards requirements.

Intermittency of PV inverters will add to the existing variations in load. Load variation is a problem that utilities already understand and manage well. But PV intermittency will increase the variation, meaning utilities require more close monitoring and control of voltage. Although this sounds like a small problem, with obvious technical solutions, the increase in cost and the change of culture and practice required will pose a big problem to many utilities.

6.1.2 Stability effects

Stability refers to the underlying balance of generation and consumption, manifested by steady and acceptable values for voltage and frequency. The frequency is generally considered such a good indication of generation and load balance that small changes can be used to attempt to avoid catastrophic collapse. In the event of a shortfall in generation, the frequency falls, triggering under frequency load shedding relays to reduce consumption. When this happens, customers are affected by blackouts in the areas where the relays operate.

Contemporary power networks are designed around steady state theory and operate by allowing large safety margins in capacity. Reserve power is made available in the expectation that increases in demand can then be accommodated quickly without overloading the network. Large reserves of spinning fossil-fuel generators are undesirable, because of both increasing financial pressures and concerns about the environment. However, excessive reduction in safety margins may increase the risk of a catastrophic failure as a result of insufficient power quality and generation capacity to resolve a fault. Cascading failure can be defined as “a sequence of dependant failures of individual components that successively weakens the power system” [67]. The initial cause may be random but subsequent events are linked by electrical, control/protection devices or human error. Disturbances that cascade beyond the local area generally do so as a consequence of unexpected protection operation [68]. This type of blackout has attracted considerable attention because of the number of customers affected, and the obvious implications about the vulnerability and brittleness of the grid. There are many examples of this type of failure, most well-known occurring in August 2003 when 50 million customers in the USA and Canada lost power for up to two days at a financial cost of $6 - $10 billion [69],[70].

Stability problems are not confined to the deployment of solar PV generation, but can be triggered by a variety of causes, including extreme weather, unexpected operation of switchgear, and lack of capacity. However, there are some disturbing reports of instability caused by wind, and as PV is likely to cause similar problems, this should be a warning of what is possible [39][83]:

“An abrupt loss of 1,200 megawatts of wind energy production on Feb. 26 [2008] caught the Electric Reliability Council of Texas (ERCOT) Inc. by surprise and forced it to declare emergency conditions”

“ERCOT said the sharp drop in production during a three-hour period - while overall electricity loads were increasing -threatened the stability of the power grid and could have caused rolling blackouts.”

“If a system can go unstable in the winter because 1,500 MW of expected wind turns into 400 MW wind and then fossil has to scramble to come online... that’s a big issue.”

This abrupt drop in wind power production, a quicker than expected evening load ramp-up and an unexpected loss of conventional generation forced ERCOT to make an emergency declaration in response to a potential major grid stability problem [83]. With a more accurate generation and demand forecast, ERCOT could have scheduled additional generation to be available in advance of the evening load pickup and avoided the need for this emergency response.

A detailed study of a large PV plant using a PSSE4 model highlights the problem of the lack of knowledge of the effects of solar intermittency on the network: “Although a few studies are found in the literature, there are no standard tools yet developed to analyze these complex scenarios. In most cases, custom models and tools are used along with some commercial programs to address the (stability) impact of large solar PV plants on electrical power systems.” [45]

These authors conclude that “with large PV plant, the system is more vulnerable to stability problems. Those scenarios can drive the system to unstable operating point if protection is not designed with proper consideration.” They also make the point that further research is required in order to understand these phenomena.

Depending on the type of network, time of year, demand scenario and unit operation and maintenance schedule, the power network can be a finely balanced system that can be triggered into a state of catastrophic collapse by an event such as that discussed above. With increasing penetration of PV, some fear that the intermittent nature of PV generation may be able to provide just the trigger that can lead to such collapse. Although network operators already manage instability, an increasing penetration of PV systems will add another potential source of instability, and one network operators are not familiar with.

6.1.3 Ramp rate effects

As discussed earlier, the rate of change of power, or ramp rate, is the amount by which the power output of a PV array changes within a given time period. If the ramp rate is too high, centralised power stations may be unable to react in time. The result of this would be voltage and frequency deviations. In extreme cases this would lead to power system instability, which could result in collapse and blackouts. In order to prevent such an occurrence, the most likely scenario is that power networks would be forced to provide rapid-peaking generation plants, such as gas, to provide voltage and frequency support. This is known as an ancillary service. Of course, storage can also be used to provide ancillary services, but this is not yet widely used. The ramp rate is also a measure of how fast ancillary services would have to react in order to stabilise the system. A small amount of intermittent generation (a small PV array) will have a smaller range of variation than a larger array, but the variation will occur over a smaller time scale than with a large array. There is evidence in the literature that as the penetration of PV rises, the ramp rate problem will be reduced, due to the averaging effect of many PV systems dispersed geographically [46]. However, there is also evidence that there may be a limit on the effect of geographical dispersion. For example a study on large scale PV sites across the American southwest showed that even spread over 280km, intermittency still requires very high ramp rate power generation [47].

Figure 55 shows ideal and actual power output from a photovoltaic array for a 24-hour period. The former is the theoretical maximum for a sunny day, while the latter is actual data taken from CSIRO’s Energy Centre at Newcastle during a cloudy day in winter.

Figure 55 An ideal generation profile for solar PV, compared with a real profile from a cloudy day in winter

Based on these data, the maximum ramp rate under ideal conditions can be calculated as 8 Watts per second. However, the maximum ramp rate for a cloudy day was calculated as 730 Watts per second.

A model has been developed based on the work of Marcos et al. [77] which allows prediction of the power output from the PV plant if the size of the PV array and the cloud characteristics are known. This model is described further in Section 11.1 of this report. The output of the model, which is the predicted PV plant output power, can be analysed to determine potential power ramp rates and the likelihood of occurrence, using similar analysis to that performed in Section 9.2. This allows estimation of the probability density function of ramp rate, which can be used to predict the effects of a particular PV array upon the local network. Such prediction should be done with a clear understanding that every PV array will be different according to its size, type, local weather, local load profiles, network type and method of connection to the network. This information could then be used to estimate the amount of fast-acting generation required to provide the ancillary services for network stabilisation.

6.1.4 Capacity factor

Apart from these obvious and immediate technical problems, there are other issues to be considered, including that the power supplied by a PV array is not schedulable, and even when stability is managed there is the issue of how power from a PV array can be traded on the electricity market. A PV array, will not always supply the rated power its nameplate size implies. The ratio of actual to potential power supplied is known as the capacity factor. There are various measures employed to characterise a PV array for its actual capacity and contribution to network generation needs. These measures can be useful in determining the level of backup generation required to offset the variability or lack of generation during periods of cloudy weather. To summarise:

  1. Effective Load Carrying Capability – a measure of how much extra load can be supported after adding PV without changing system reliability
  2. Load Duration Capacity – an average output of PV for the peak of the load duration curve
  3. Demand Time Interval Matching – the difference in peak demand with and without PV
  4. Time Window Season – probability of providing a minimum power output within a given time window
  5. Relative measures of intermittency based on Fourier spectra or fractal measures within a given time scale

Using one of these measures an effective MWh can be calculated. This will be lower than total MWh served, in order to quantify the uncertainty. If a PV array contributes a total MWh and an effective MWh, the ancillary services cost can be charged as the difference between these. In other words, how much ancillary service is needed to make the effective MWh into the total MWh served? The more intermittent the PV, the bigger the difference between effective MWh and MWh served, the bigger the share of ancillary services for which it is charged will be.

The capacity factor for wind power is considered to be between 1/6 and 1/3. At best case, 3 GW of wind needs to be installed to generate total annual energy equivalent to that from 1 GW of conventional generation [46]. According to the same paper (citing the author’s own earlier work) Germany had to build nearly 50% reserve margin to provide back up for its wind generation. There is little reason to suppose solar PV will be any different. Wind projects in the USA are estimated to have a capacity factor of 40% at best, while coal fired has 60% and nuclear power over 90%. Solar PV is estimated to have a rapidly rising capacity factor currently at about 21% [46]. It cannot be assumed that the success of European countries such as Germany and Denmark in integrating renewable energy will be matched in Australia. Australia has a very different network structure, characterised by low populations and large distances, resulting in weak networks compared with the US and Europe.