12 Summary of key findings
The current state of worldwide research on renewable generation intermittency was summarised in the early sections of this report. One of the main challenges to the power system is associated with the instantaneous penetration of intermittent solar generation. As solar generation is viewed as negative load, when this is combined with the actual system load, the characteristics of the resulting net load which has to be supplied by other generating resources in the system changes. Knowing what kind of variability to expect when high penetration solar power is integrated into the system is important for forecasting and planning purposes. Of concern to utilities is when load and solar power move in opposite directions at the same time, creating large changes in the net load. Various studies have shown that a high penetration of intermittent generation results in greater variability in the net load compared with the variability in the original load alone without solar or wind.
There is very little published literature to be found discussing observed impacts of high penetration solar intermittency. The majority of work discovered focused on modelling impacts rather than actual observation of impacts. Studies have shown that adequate system flexibility is a key requirement for managing increased levels of intermittent renewable generation. As a result, conventional generators are forced to be more flexible with their output resulting in a higher per unit cost. A study carried out in Gardner, Massachusetts, shows how rapidly the net load of a system can vary significantly, which is likely to put added pressure on conventional generating resources on the system to vary their output rapidly. A good example of output variations that can be expected from a large-scalesystem can be seen from the output of a 4.6 MW PV system located in Springerville, Arizona, where large abrupt power output drops, from about 4000 kW to 500 kW, were seen to occur over extremely short timeframes. The literature indicates that cheaper less flexible plants will need to be replaced with more flexible expensive plants to accommodate high penetrations of solar generation. Otherwise, a significantly larger amount of ancillary services or additional generation would be required to manage PV power output fluctuations.
Some general conclusions that can be drawn from these studies include:
- The amount of solar generation that can be integrated into the utility power system without compromising and reliability varies widely. The determining factors are the amount of a utility’s load fluctuation and the regulating capability of existing conventional generating units. This observation indicates the effect of solar generation intermittency on the power system is not uniform and is case sensitive. Hence, a general cause-effect conclusion cannot be drawn.
- Although high penetration levels of solar generation have the potential to cause adverse network impacts, corrective measures such as assigning more generating units to regulating duty or installing fast-response combined-cycle generators are available. These measures can be effective if carefully planned. High penetration limits have been shown to be possible (in simulations) after adding fast-response, combined-cycle generating units to the existing generation mix. Another method is to control the solar generation output under intermittent cloud coverage during periods of peak system demand when the network has fewer generating units on standby and less on-line regulating capacity. These corrective measures, however, may cause the system to deviate from its optimal operating condition, thus adversely affecting the economics of solar generation.
It is widely agreed that accurate forecasting is an essential element for the successful integration of large amounts of intermittent solar generation and for solar power to be economically viable. Forecasting at various timescales is required. More accurate day-ahead prediction of renewable resources is required for more accurate unit commitment. Short timescale predictions are also needed (which could be obtained by tracking cloud movements), while numerical weather models can be used to predict insolation out to a number of days.
Power output from both solar and wind generating sources are known to vary considerably with varying irradiance and wind speeds respectively. Analysis of the results illustrated in the studies carried out on wind and solar variability indicate that the output of individual wind turbines and individual solar arrays are similarly variable in nature in the second-to-second timeframe. When wind and solar output are taken in aggregate and analysed at 1-minute to 10-minute intervals, wind seems to benefit more from the smoothing effect associated with aggregation, showing less variability than solar. When behaviour is observed at 1-hour intervals, again in aggregate, the opposite is true and wind is shown to be more variable.
When considering the variability of PV plants versus CST plants, a significant difference is evident when looking at partly cloudy day output profiles, where the variability of PV plants is far greater than for CST plants. This is primarily because CST plants have more thermal inertia in their underlying equipment which acts as a buffer to avoid immediate drops in the temperature due to passing clouds, thereby contributing to significant reduction in variability.
Existing literature was found to contain conflicting outcomes, possibly due to the lack of quality data, and consequently often overemphasised anecdotal evidence.
Table 25 lists two examples of conflicting outcomes reported. By looking at the outcomes of the studies that report the comparison of solar and wind power variability, one could deduce that the variability of both solar and wind resources is dependent upon factors such as geographical location, spatial diversity and size of the renewable generation system.
Discussions at recent IEA Task 14 workshops indicate there has been substantial growth worldwide in generation from intermittent renewable resources, mainly PV and wind. Countries involved in Task 14 including Germany, Denmark, Spain,and are currently experiencing a high rate of increase in PV installations, both small- and large-scale systems, and are expecting to see an increase in their intermittent generation capacity by a few hundred percent by 2020. The Task 14 representatives acknowledged that an area of considerable challenge they expect to face with increased PV penetration is to do with the variability and predictability of solar power systems. Issues reported by various Task 14 representatives include large abrupt changes in generation output as a result of intermittency, increase in net load variability on a feeder upon installation of a large PV plant, and frequent occurrences of high ramp-rate variations in PV output power.
|Study outcomes||Corresponding conflicting outcomes|
|A study by theIndependent System Operator (NYISO) reported that significant cost savings could be achieved by integrating intermittent generation (wind in this case) due to the displacement of fuel, primarily natural gas, by wind and by having accurate forecasts.||The POVRY study for Europe reported that the volatility introduced by the highly variable wind and solar output will directly impact on cost. Prices are expected to become peakier and less predictable, representative of the nature of weather systems.|
|Analysis performed in a Western US study found that savings can be achieved from the introduction of solar power through the displacement of gas and coal fired power generation and a price on carbon. Large forecast errors would, however, cause expensive generation to be brought online.||Simulations performed in the Texas, US (ERCOT) grid where different mixes of wind and solar power were modelled reported that high penetration of intermittent generation will increase system costs due to the upgrade of conventional generation equipment required to achieve increased system flexibility.|
|Analysis of results in a study carried out by theon the Independent System Operator (CAISO) system showed wind to be more variable than solar when both wind and solar power were considered in aggregate at 1-hour intervals.||It was reported in a Swedish study that the smoothing effect due to aggregation is greater for wind than for solar at the hourly timescale, which conflicts results seen in the CAISO study.|
In order to investigate the issue of solar intermittency in the Australian context, an industry workshop and online survey were conducted by the project team to understand the issues key solar industry stakeholders are facing due to solar intermittency, and their perspectives on what is needed to remove barriers caused by the intermittent nature of solar. The key findings are:
- The impacts of intermittency from large-scale and small-scale solar systems on the electricity network need to be studied separately. Large-scale solar systems are often located in remote areas where the environmental conditions are noticeably different to urban areas in where majority of the small-scale solar systems are located. The grid configuration and strength are also different in remote and urban areas; hence the impacts of solar intermittency on the electricity network are likely to be different.
- The ramp up and down rates of solar, more for PV than CST, are potentially significantly higher than wind due to the lack of inertia. It is important to investigate the ramp rates and time-frames of intermittency in order to determine how quickly systems have to respond.
- Solar output is more predictable than wind in the very short term as the movement of clouds is visible, but accurate prediction for long-term solar output is required in order to determine ways of effectively compensating for solar intermittency. Accurate solar forecasting with high temporal resolution is required to manage solar intermittency issues.
- Higher resolution data than is currently available is required to study solar intermittency and its impacts on stability of the grid. High resolution solar data from both large numbers of small-scale solar systems aggregated and large-scale solar systems is required to investigate the effects of various temporal variances on the Australian electricity network.
- Solar intermittency will likely cause power quality issues which need to be investigated. Sudden shadows by clouds appear to produce more rapid flicker than sudden wind changes.
- Possible rates of change of power and the performance of the network due to solar intermittency need to be investigated in order to determine the type of ancillary services required and to determine whether existing mechanisms are sufficient for intermittency compensation purposes.
Australian solar industry stakeholders have expressed considerable interest in further examining the issues detailed above and are keen to learn more about the likely impacts of solar intermittency on Australian electricity networks with an increasing penetration of both small- and large-scale solar systems.
To demonstrate the effect of rapid fluctuations in solar radiation and to understand the variability of solar power plants, ten months of high-resolution 10-second data from the Desert Knowledge Australia Solar Centre (DKASC), Australia, was collected and analysed. The data comprised the total diffused horizontal irradiance, total output power of the PV arrays at the Solar Centre and line-to-neutral AC voltage. To study PV power output ramp rates for a small-scale system and also a large-scale system in Australia, data was also collected and analysed for a 22 kW PV system at the CSIRO Energy Centre in Newcastle with 5-second resolution and for a 1.22 MW PV system at The University of(currently Australia’s largest flat panel PV system) with 1-minute resolution.
Analysis of the data collected has been conducted to evaluate the occurrences of power fluctuations over time for the DKASC. The raw data was processed and occurrences of continuous ramping of the solar plant’s output power for periods of ten, twenty, thirty, forty, fifty and sixty seconds were extracted and studied. Analysis of 10-second ramp events resulted in a total of 422 occurrences of variations in PV plant output power of at least 60kW (31% of plant rating) in ten seconds. 7 occurrences of output power drops of between 130kW and 140kW (66% and 71% of plant rating respectively) in just ten seconds were also recorded. A significant number of events where the output power was seen to vary by more than 77% of plant rating over timeframes of twenty, thirty, forty, fifty and sixty seconds were also recorded at the DKASC. The data for line-to-neutral AC voltage measured at the switchboard at the DKASC were also analysed. A positive correlation between the variations in irradiance, output power and voltage was observed. A larger number of ramp-down events with higher ramp rates were observed compared with ramp-up events for this system.
Analysis of the data for the 1.22 MW PV system at UQ showed 21 events of output power variations greater than 720 kW (59% of plant rating) in 1-minute periods in the five-month period in which data was collected. Three occurrences of output power drops greater than 900kW (74% of plant rating) in five minutes and another three occurrences of power drops greater than 900kW in six minutes were also recorded. This analysis demonstrates the possible significance of large fluctuations of PV power over very short timeframes to grid operation. In a high penetration scenario, these fluctuations could potentially cause large variations in instantaneous net load causing conventional generators in the power system to vary their output significantly very quickly. This shows the importance for a power system to have greater flexibility in order to accommodate high penetration of intermittent solar generation.
A simulation model developed at CSIRO was used to examine the likely impacts of output power fluctuations seen at the DKASC on various types of Australian electricity networks with different penetration levels of solar power. The effects of intermittency on voltage and network stability were modelled and observed for four different scenarios, and are summarised as follows. Note that load variations were not considered in this study.
- Strong (e.g. urban and relatively large) grid with low penetration solar power (10% PV penetration with 0.01p.u. lumped grid impedance): This type of grid situation would probably only apply in an urban area close to a substation. Results show voltage variation of only 0.14V.
- Weak (e.g. remote and relatively small) grid with low penetration solar power (10% PV penetration with 0.2p.u. lumped grid impedance): This is to simulate a rural grid under Australian conditions with a low level of PV penetration. The voltage was seen to sag as power from the PV array drops and the range of voltage variation observed was 0.8V. This variation of voltage would probably go unnoticed by customers.
- Strong (e.g. urban and relatively large) grid with high penetration of solar power (40% PV penetration with 0.01p.u. lumped grid impedance): The voltage is seen to change very little, with a range of 0.35V. This is because the strong grid acts as a buffer for variations in PV generation.
- Weak (e.g. remote and relatively small) grid with high penetration of solar power (40% PV penetration with 0.2p.u. lumped grid impedance): This is to simulate a long feeder, typical of Australian conditions. Results show voltage swings of 4V, similar to the actual data recorded from the DKASC array over a 15-minute period.
This shows that when solar generation, mainly PV, is attached to a rural feeder where the grid has high impedance, an increase in penetration is likely to cause large voltage swings to be observed. These increased voltage swings may have the potential to impact the stable operation of the network. It is also likely that PV inverters would trip off with these voltage swings, resulting in larger power fluctuations and therefore worsening the voltage swings.
The larger the PV plant, the longer it takes for a cloud cover spread to shade the entire field. An analytic model has been developed where the PV plant power output is described as the signal output of a first order low pass filter whose input signal is the irradiance signal. The pole value of the filter is a function of the PV plant area. Frequency domain analysis using Discrete Fourier Transform (DFT) was applied to both the irradiance and PV plant output power data recorded during the ten-month period. The cut-off frequency for the PV plant at the DKASC was determined to be 0.0055Hz which agreed well with observations made in existing literature.
The developed model can be used to predict the power output of an existing or proposed PV plant via simulation for measurements of a given time irradiance time series. This would allow estimation of the probability density function of output power ramp rates which can be used to predict the effects of a particular PV array upon the local network. The actual power output of the PV plant at the DKASC was compared with the power output predicted by the developed model using irradiance signal as its input. The similarity between the actual and predicted output power was clearly observed, indicating the validity of the simulation model developed.