3.2 Impacts/observations: high penetration solar intermittency

The impacts of PV power plants are associated with voltage profiles, electrical losses, power factor, capacity planning, power quality, system operations and protection. Currently utility-scale solar PV plants have nominal capacities that are compatible with distribution substation MVA ratings.

During normal operations, energy market operators control and dispatch conventional generators to minimise the cost of producing electricity while maximising system reliability. Each generation unit is committed and loaded according to its heat rate, fuel cost and availability, associated transmission losses, and output ramp rate, to satisfy the electricity demand reliably at the lowest possible cost [44]. However, unlike operators of conventional generation units, renewable generation system operators have no control over the availability and quantity of solar and wind resources, as weather variations dictate the generation output of these units. The inclusion of intermittent energy technologies in the system means conventional generators not only must follow the usual load demand changes, but also make up for the output variations caused by intermittent generators. Other techniques that can be used to compensate for the output variations of intermittent generators include energy storage, load response and curtailment of generation from intermittent sources. Five normal functions of the generation operations that could be affected are:

Load-frequency control: When load exceeds generation, the system frequency will drop, and vice versa. When a change in the system frequency is detected, power system operators increase or decrease the output of conventional generators to match the load. Intermittent renewable generation technologies generally have not participated in system frequency regulation and have output power that is independent of system frequency. However, through output curtailment, there is no engineering reason why renewable generation could not participate in the FCAS1 Lower markets of the Australian National Electricity Market (NEM). An example of this is Germany, where the grid operator has the capability of ‘turning down’ PV output for large PV plants if required.

Load following: If an increase in solar or wind power does not coincide with a system load increase, other generating units in the system will have to be off-loaded so as to utilise the solar or wind power while keeping the system balanced. When meteorological conditions cause a decrease of solar or wind power, output from other units has to increase to take up the generation slack. Networks normally use intermediate plants to follow the load. The integration of high penetration intermittent generation may result in increased load-following duties for the conventional generators assigned for system regulation.

Ramping rate: Ramping rate represents the generator’s ability to change its output. The ramping rate of on-line generators has to be able to follow the combined load changes and output fluctuations of intermittent generators when intermittent generation is added to the electricity system; for example, load increase and intermittent generation decrease simultaneously, or vice versa.

Unloadable generation: The down-ramping rate of a generator may be different from its up-ramping rate. Both are important to meet the normal system load-following requirement. The amount of generation that can be off-loaded (down ramping) is called unloadable generation. In order to accommodate the maximum output from intermittent generating technologies, system operators have to make certain that on-line conventional generators can be backed down quickly enough, particularly when facing a simultaneous sudden increase of intermittent generation output and a reduction of system load. Such an accommodation to absorb energy from intermittent generation cannot be made by tripping off a unit because the unit may be needed again soon after being taken off-line.

Operating reserve: The impact on the electric system operating reserve is also related to the intermittency of solar and wind generation technologies. Operating reserves are maintained to guard against sudden loss of generation and unexpected load fluctuations. Any load and generation variations that cannot be forecast have to be considered when determining the amount of operating reserve. Carrying operating reserves is expensive. If the short-term fluctuations of intermittent renewable generators cannot be predicted accurately, more operating reserves will have to be scheduled to ensure the system can be kept regulated within standards. This requirement will increase the cost of integrating intermittent solar and wind systems.

These five potential impacts on operational requirements mean an increased proportion of conventional generation units may need to be brought on-line or put on regulating duty to manage higher penetration intermittent renewable generation, which may increase the system operating cost.

One of the objectives of sub-task 3 of the IEA Task 14 Photovoltaics Workshop is to look into the impacts of high penetration solar on distribution grids. Presentations at recent meetings in Colorado and Lisbon included case studies showing impacts which give reason for concern.

The issues of concern include:

  • fast ramp times in PV output due to cloud activity making voltage and frequency regulation difficult
  • displacement of conventional generation providing ancillary grid services
  • aggregate loss of PV during faults and contingencies (due to under-voltage and under-frequency)
  • insufficiently accurate forecasting making scheduling difficult.

According to [13], conventional generators are forced to be more flexible with their output, resulting in a higher per unit cost. Adequate system flexibility is a key requirement for managing increased levels of intermittent renewable generation; Section 3.4 discusses how flexibility requirements can be met.

This section presents observations of actual case studies of high penetration PV intermittency. Measured data was given for a report on a PV plant in Gardner, Massachusetts and a small scale study done in Italy. Unfortunately the majority of work discovered is modelling of impacts rather than observation of impacts.

Cloud activity is the main reason for variability in PV output. High frequency sampling (1-second) at the Gardner, Massachusetts [14] site shows the kind of ramp rates possible in PV output and net load due to passing clouds. Figure 13 shows the impact of a passing cloud on the irradiance and net load of the site, which consists of 53 residential properties and 28 PV installations. The irradiance is seen to initially dip sharply just after 14:13, causing the net load to increase sharply as the cloud edge begins to move across the site. During the one minute and thirty seconds period between the onset and completion of the cloud passing, the net load is seen to increase by 45kW from a condition where the site was exporting 10kW to one of importing 35kW. At 14:15, the irradiance is seen to increase from being near zero to 85% in 40 seconds, and the site is once again exporting power. This case study shows how rapidly the net load of a system can vary significantly.

Figure 13 Irradiance and net system power output during cloud passage [14]

Figure 14 shows the site power flow (P34) over a 6-minute period with passing clouds. Even taking into account the smoothing effect aggregation has on PV output, large drops in power output are observed, including a drop from 50kW to 8kW in approximately one minute.

Figure 14 Total site power and PV generation over a 6-minute period [14]

Large variations in PV output for systems with high penetration PV will result in proportionally large variations in net load. Existing methods for managing large scale intermittent generation are not expected to be sufficient. Again, Section 3.4 contains further discussion on how large scale intermittent generation may be managed. A good example of output variations that can be expected from a large-scale PV system can be seen from the output of a 4.6MW PV system located in Springerville, Arizona, US. The extent of intermittency exhibited by such a large system can be seen in Figure 15, where the PV output data was sampled every ten seconds. Large abrupt power output drops, from about 4000 kW to 500 kW, can be seen to occur over extremely short timeframes.

Figure 15 Power output of a 4.6 MW PV system on a partly cloudy day in Arizona [39]

A ‘near-occurrence’ of instability caused by intermittency of wind generation is when ERCOT2 was forced to declare emergency conditions when an abrupt loss of 1,2000 MW of wind energy production caught them by surprise on 26 February, 2008 [39]. The sharp drop in production occurred during a three-hour period when overall electricity loads were increasing. This threatened the stability of the power grid and had the potential to cause rolling blackout. Large-scale PV penetration may cause similar problems due to significant magnitude and ramp rates of power output variations that can be seen to occur, an example being the case shown in Figure 15.

Installed PV capacity in Germany has increased greatly in recent years. By the end of 2010 approximately 80% of cumulative installed PV capacity (about 14 GW) was connected to the low voltage network [40]. Before a transitional arrangement was introduced by the VDE|FNN e.V. in April 2011, low voltage generation plants were required to be switched off immediately if system frequency increased to 50.2 Hz. In a worst case scenario, up to about 9,000 MW of power from PV systems would disconnect from the network if system frequency increased to 50.2 Hz [40]. That would cause a large instantaneous loss of PV power on the network and a sharp increase in the load seen by centralised generators. This in turn could affect the stability of the grid. Reaching a system frequency value of 50.2 Hz during normal operations is as yet quite unlikely but any unexpected large-scale disturbance followed by an abnormal system condition would pose significant risks and cause the system frequency to increase due to an oversupply of electrical power. An example of this is the European power grid failure in 2006 due to power imbalance [41] and the blackout in Italy in 2003 [42]. In both cases, Germany belonged to an exporting network region in which the frequency value increased to 50.2 Hz. The European grid is designed only for a sudden loss of 3,000 MW of generating capacity. If similar disturbances were to occur on sunny days with the current PV capacity during high supply from those PV systems, their power infeed would be lost. On sunny days the current PV capacity in Germany exceeds the current maximum value of 3,000 MW by several times. As a result, there would be a high probability of a large-scale failure to the electrical supply in those parts of Europe affected by this phenomenon. Note that the 50.2 Hz issue in Germany is not specifically an intermittency issue, more one of inverter or regulatory parameter settings.

The power quality provided by a photovoltaic system is described by the voltage, frequency, harmonics of voltage and current, flicker and power factor [15]. An experimental analysis aimed at evaluating the effects of many PV plants on the power quality of a grid was performed in [15]. In this case, it was found that the Total Harmonic Distortion (THD) of the current injected into the grid, independent of grid characteristics, may at times exceed standard limits. Moreover, the THD for voltage could exceed standard limits in the case of high impedance networks, typical at the distribution level. These findings were based on systems ranging from 16 to 40 kW. The penetration level of these systems is not mentioned, but considering their size it is unlikely to be large. Based on the findings of this report, higher levels of distortion are likely to be observed with increased penetration, especially in grids with relatively high impedance (suburban and rural grids). It is important to note that these results were unlikely due to harmonics caused by the PV inverters (limits for harmonic distortion in inverters, as set in Australian Standard AS4777, are significantly lower than those allowed for loads), but rather that by partly meeting the real power requirements (reducing the fundamental frequency component) of the network segment, the harmonics due to loads become more apparent. Consequently, while there is unlikely to be any significant difference in the total harmonics due to high PV penetration levels, the reduced real power requirements can lead to an increased THD ratio.