9.2 Intermittency data analysis - DKASC 196 kW PV system
As the output of any solar system is insolation dependent, the variability of generated power due to uncontrollable intermittent solar irradiance needs to be compensated to maintain system balance. To determine the regularity and rate at which generated output power can change for a local solar power plant, ten months of data (October 2010 to August 2011) was collected at a sample rate of 0.1Hz from the DKASC. The data collected and used for this analysis comprised total diffused horizontal irradiance, corresponding output power and line-to-neutral AC voltage. The various solar technologies at the DKASC combined realise a 196kW Figure 82 shows the output power (upper plot) and irradiance (lower plot) over the ten month period. A reduction in the amplitude of irradiance is seen for the second half of the ten month period which corresponds to the onset of winter months in 2011. The corresponding power output does not show reduced amplitude. This is likely due to the tracking of the solar panels and the fixed nature of the pyranometer used to measure irradiance. Another factor may be the tilt angles of the various fixed panels which may be such that they result in a relatively even power output over the year, as opposed to peak output in summer or winter.system.
Figure 83 shows the irradiance and corresponding power output at the DKASC on a partly cloudy day in May 2011. A strong positive correlation can be seen between the irradiance and corresponding power output of the PV system. Figure 84 shows the power spectrum of the fluctuations from this system over the ten month period using ten-second resolution data. The y-axis, |Y(f)|, represents the relative magnitude of the different cycles. As expected due to daily periodicity, the largest peak is seen at 24 hours (1.15 x 10-5 Hz), followed by peaks at 12 hours, 8 hours, 6 hours and so on.
9.2.1 Ramp rates for different timescales
Analysis has been conducted to evaluate the occurrences of power fluctuations over time for the DKASC. The power fluctuations were analysed using ramp rates as a measure, defined as the amount by which the power output of a PV array changes within a given time period. An irradiance increase leads to an increase in the plant output power (ramp-up). This is due to clouds leaving after previously shading panels or the sun rising at the start of a day. An irradiance reduction causes a drop in the plant output power (ramp-down), due to arrival of clouds shading the panels or the sun setting at the end of a day. High ramp-up and ramp-down events can also occur due to inverters tripping or reactivating. These events are also included in our analysis. Output power ramp events over time periods ranging from 10 seconds to 60 seconds were extracted from the raw data obtained from the DKASC. To obtain ramp rate distributions for timescales of twenty, thirty, forty, fifty and sixty seconds, instead of downsampling the 10-second data, ramp events and ramp rates for every interval of the different timescales were captured, meaning sampling intervals overlap through the dataset and no samples are neglected. For example, a 60-second block of data contains five 20-second ramp events8 instead of three if the data was downsampled. The output power ramp rates for various timescales, At, were calculated as the difference, AP, between the power values at each end of the time interval divided by At, that is:
Analysis of the data showed there was a total of 8,072 occurrences of ramp-up rates of 2kW/s (1% of plant rating per second) or more for ten-second periods over the ten months. Figure 85 shows the 120 of these occurrences that were over 6kW/s which corresponds to increases in output power of at least 60kW (31% of plant rating) in ten seconds. The analysis also showed there were 8,310 occurrences of ramp-down rates of more than 2kW/s or more for ten-second periods throughout the ten months. Figure 86 shows the 302 of these occurrences which were for ramp rates of 6kW/s or more. Ramp-down rates of between 13kW/s and 14kW/s were observed to occur seven times, corresponding to a power drop of between 130kW and 140kW (66% and 71% of plant rating respectively) in just ten seconds.
Occurrences of continuous ramping of the solar plant’s output power for periods of twenty, thirty, forty, fifty and sixty seconds were also analysed. Figure 87 shows occurrences of ramp-up rates for variations recorded in twenty seconds. A total of 5,946 occurrences of output power ramp-up events with rates of at least 2kW/s over 20 seconds were recorded. The corresponding total number of ramp-down events for 20-second periods with ramp rates of more than 2kW/s was found to be 6,298, out of which 768 were fluctuations with rates of 4kW/s or more, that is, a drop in output power of 80kW (41% of plant rating) or more in 20 seconds. This can be seen in Figure 88. Three occurrences of ramp-down events with rates higher than 8kW/s were recorded which corresponds to a drop in output power of more than 82% of the plant rating.
The distribution of ramp-up event occurrences over 30-second periods with rates greater than or equal to 3kW/s is shown in Figure 89. There were 862 such events in total, including 56 occurrences of ramp-up events with rates greater than 4kW/s corresponding to output power increases of more than 120kW (61% of plant rating) in half a minute. Figure 90 shows the distribution of output power ramp-down events over 30-second periods for rates greater than 3kW/s, the total number of such events was found to be 1,209. The number of times the output power dropped at least 150kW (77% of plant rating) in half a minute was found to be 37.
Figure 91 shows the distribution of ramp-up event occurrences over 40-second periods with rates greater than 2.5kW/s, corresponding to output power increase of at least 100kW in 40 seconds. There were 218 events of increase in power output of at least 120kW (61% of plant rating) in 40 seconds recorded. Figure 92 shows the distribution of output power ramp-down events over 40-second periods for rates greater than 2.5kW/s and five events of output power drops of more than 160kW (82% of plant rating) in 40 seconds were recorded.
Figure 93 shows occurrences of various ramp-up rates for variations recorded in fifty second periods. A total of 1,344 occurrences of ramp-up rates of 2 kW/s or more were observed which corresponds to increases in power output of at least 100kW (51% of plant rating) in fifty seconds. The corresponding total number of ramp down occurrences for 2kW/s or more for fifty seconds is 1,515, of which 54 were for ramp-down rates of 3kW/s or more (greater than 77% drop in plant power output in fifty seconds), as can be seen in Figure 94.
The distribution of ramp-up event occurrences over 60-second periods with rates greater than or equal to 1.5kW/s is shown in Figure 95. The total number of such events was 2,308, including 18 occurrences of ramp-up events with rates greater than 2.5kW/s corresponding to output power increases of more than 150kW (77% of plant rating) in 60 seconds. Figure 96 shows the distribution of output power ramp-down events over 60-second periods for rates greater than 1.5kW/s, the total number of which was found to be 2,163. The output power dropped at least 150kW in 60 seconds 56 times.
The number of occurrences of ramp-up events over various time periods is summarised in Table 18. It can be seen that the majority of the fluctuations observed over all time periods are small variations with ramp rates of less than 1kW/s. A significant number of rapid increases in output power were observed with ramp rates of more than 5kW/s over 10- and 20-second periods which correspond to increases of more than 50kW (26% of plant rating) and 100kW (51% of plant rating) respectively in a short time frame. An illustration of the distribution of ramp-up events over time periods ranging from 20 seconds to 60 seconds is shown in Figure 97.
Note ‘N/A’ refers to ramp-rate events that are not feasible - such ramp rates over those timescales will exceed the capacity of the plant.
Table 19 presents a summary of occurrences of ramp-down events over various time periods. Similar to ramp-up events, the majority of the total number of fluctuations over all time periods were small variations with ramp rates of less than 1kW/s. However, a larger number of ramp-down events with higher ramp rates were observed compared to ramp-up events. A significant amount of rapid drops in output power with ramp rates greater than 5kW/s, corresponding to at least 50kW (26%) and 100kW (51%) for 10- and 20-second time periods respectively, were observed and this is greater than that of corresponding ramp-up events. Figure 98 illustrates the distribution of ramp-down events over time periods ranging from 20 seconds to 60 seconds.
This analysis demonstrates the possible significance of rapid fluctuations of PV power to grid stability. This effect can obviously be mitigated to some degree by distributing the PV systems over wider areas in order to de-correlate system behaviour.
Note: ‘N/A’ refers to ramp-rate events that are not feasible - such ramp rates over those timescales will exceed the capacity of the plant.
Data from a concentrating solar thermal (CST) plant could not be obtained for this project for the purpose of intermittency timescale and ramp rate analysis. However, depending on the type of CST technology used, CST systems may be less susceptible to rapid ramp-up or ramp-down situations due to inherent thermal inertia in the system, and short-term cloud event ride through is believed to be achievable. Figure 99 shows the output from PV and CST plants without dedicated storage on a cloudy day. A significant number of rapid fluctuations are seen in the output power of the PV plant while the output power of the CST plant is fairly steady throughout the main part of the day with significantly slower changes.
Source: Mehos et al.,Power & Energy Magazine, May/June 2009
9.2.2 Voltage variations
Data of average line-to-neutral AC voltage of the three phases of the DKASC with 10-second resolution were also obtained for a period of ten months, from October 2010 to August 2011. The voltage readings were taken at the centralised switchboard using a class 0.5 ION 7550 energy meter which monitors the main feed to the switchboard. This meter is located at the output of the PV plants. The DKASC PV system is connected at the end of a feeder to which no loads other than a few street lights are connected to. The main feed is approximately 350 metres long, 3-phase+neutral, connected from a 22kV step-down transformer. The 22kV substation is the first substation on the feeder into the Desert Knowledge precinct which becomes a 22kV ring supplying about 16 different buildings, comprising offices and teaching spaces. Figure 100 shows a plot of the voltage data over the ten month period. The maximum voltage recorded was 269.39V.
Figure 101 shows the distribution of voltage measured at the DKASC. For the analysis conducted in our study, voltage drops to zero (representing the loss of main grid supply) were ignored, focussing instead on the system during normal operation. The average line-to-neutral voltage of the three phases was found to be greater than 250V for 79.8% of the time over the ten months, and the proportion that was recorded to be over 253V (upper limit of grid voltage, i.e. 230V +10%) was 29.4%. The tap setting over the one-year period for which data was analysed was high (433V), resulting in observed voltage of about 250V almost all the time. The tap setting has been reduced to 415V in late November 2011 for voltage to be around the 240V level. As there was no end consumer in this case, the high voltage had no effect on any load: there was none apart from some street lights. The trip settings of the PV inverters are set higher than what the Australian standard AS4777 allows, in agreement with the local utility, i.e. a bigger voltage window.
Figure 102 shows the irradiance, output power and corresponding voltage recorded at the DKASC over a seven day period from 7 to 13 February 2011. It can be seen that all seven days experienced a significant number of passing clouds, causing many spikes in the generated power output during the course of the week. A similar plot for the duration of one day is shown in Figure 103. To illustrate the impacts of clear and sunny days on the PV plant output power and voltage, Figure 104 shows the effects of passing clouds on the plant output power and voltage on two consecutive days, 19 (cloudy) and 20 (clear) December 2010. A positive correlation between the irradiance, output power and output voltage can be seen from Figure 102, Figure 103 and Figure 104. In Figure 104, a spike in the output power and voltage waveforms can be seen just after midday on a clear day, on 20 December 2011. This is likely due to the tripping of inverters, causing spikes in both the output power and line-to-neutral voltage.
To study the voltage behaviour at the DKASC plant, the voltage and irradiance over one day were plotted together for comparison, as shown below in Figure 105(a), with the voltage and power output for the same day shown in Figure 105(b). A positive correlation is again observed between the voltage, irradiance and output power of the plant. Fluctuations in the voltage seen outside daylight hours are likely due to variations in loads connected to adjacent feeders. The magnitudes of these fluctuations are, however, not as large as those observed during daylight hours due to passing clouds.
Similar plots to Figure 105 but restricted to a 50-minute window around midday on 7 February 2011 can be seen in Figure 106. Figure 106(a) compares the voltage and insolation fluctuations over the period, while Figure 106(b) compares the voltage and plant output power fluctuations. A strong positive correlation is again observed between all the three parameters.