In the June “Renewable California” post I presented a brief analysis of California’s progress towards its goal of cutting greenhouse gas emissions at least 40 percent below 1990 levels by 2030 based on annual generation data. Hourly grid data for the period between April 20, 2010 and March 9, 2016 are now available, and this post reviews them to see what they add. The conclusion is basically the same as before – that despite all the legislation that California has passed in an attempt to stimulate the growth of renewables the state has not progressed at all. The percentage of renewables in California’s energy mix is still about the same as it was in 2010 and the percentage of low-carbon generation in the mix has decreased slightly. The California “Duck Curve” also remains a matter of concern.
The data used in this post are from the California Independent System Operator Corporation (CAISO), which combines grid data from electricity producers in California. They were reportedly collated by Todd D. and linked to in a comment by Thinkstoomuch in the June 2016 “Renewable California” post, so a hat tip to these two gentlemen. The data are contained in this XLS spreadsheet.
Daily average data plots:
The CAISO data are provided at hourly intervals so the spreadsheet contains over 50,000 lines. Plotting the data for the entire 6-year period would have generated a very messy graph, so I calculated daily means and plotted them instead. The results are shown in Figure 1.
Figure 1: Daily average generation by source, April 20, 2010 to March 9, 2016
The first thing one notices is how little things have changed over the last six years. Demand (assumed to be equal to total generation) has been flat over this period. Thermal generation and imports have remained substantially the same. Renewables generation has increased, but this increase is offset by a decrease in hydro generation, which CAISO puts in a separate category. The decrease in nuclear generation caused by the shutdown of the San Onofre nuclear plant in early 2012 is, however, clearly visible. Also evident is the seasonal lag between hydro generation, renewables generation and imports, which peak in or around June, and demand, which peaks in or around September.
We will now take a look at individual generation sources. Figure 2 plots thermal generation, which is effectively all natural gas (see Figure 4 in the “Renewables California” post). Thermal generation increased to pick up the slack after the San Onofre shutdown in 2012 but since then it has been flat:
Figure 2: Daily average thermal generation, April 20, 2010 to March 9, 2016
Figure 3 plots imports from out-of-state power plants, some of which are wholly or partly owned by California utilities and some of which are not. The generation mix is not broken out, but a significant proportion of it is believed to be coal. Imports tend to peak around mid-year, two or three months before the demand peak, probably because hydro generation from the Bonneville dam – another significant source of California’s imports –peaks at this time:
Figure 3: Daily average imports, April 20, 2010 to March 9, 2016
Next comes renewables. It’s a commentary on the California mindset that while imports are not segregated by source renewables are segregated into no fewer than seven categories (Figure 4). We see no growth in geothermal, biomass, biogas and small hydro and only minor growth in wind. Only solar PV has grown substantially. (Solar thermal is insignificant. I had to color it black to make it visible.)
Figure 4: Daily average renewables generation by source, April 20, 2010 to March 9, 2016
An interesting feature of Figure 4 is that wind generation peaks around mid-year and decreases to near-zero in the winter, just like solar only more pronounced. According to the US Energy Information Agency (EIA) this is because sea breezes in California blow strongly in the summer but die in the winter and because the major wind farms are located in mountain passes close to the coast where the sea breeze effect is amplified.
An even more interesting feature of the plot is that for reasons best known to the state of California large hydro (50MW or more) is not considered to be a renewable resource. Figure 5 shows what Figure 4 looks like when large hydro is counted as renewable generation, which it unquestionably should be. Now it’s difficult to see any growth in renewables at all:
Figure 5: Daily average renewables plus hydro generation, April 20, 2010 to March 9, 2016
Another source of power is also ignored. California can measure its progress towards cutting greenhouse gas emissions only by taking all low-carbon generation sources into account, and these include nuclear. Figure 6 accordingly adds nuclear generation to the Figure 5 data. Now the percentage of low-carbon generation in California’s energy mix shows an overall decrease since 2010:
Figure 6: Daily average renewables plus hydro plus nuclear (equals total low-carbon) generation, April 20, 2010 to March 9, 2016
Figure 7, which plots the percentage of California’s total generation contributed by the main generation sources in each year, summarizes the above results, with the black line defining the contribution of low-carbon sources. The key event was clearly the shutdown of the San Onofre nuclear plant in 2012, which based on the rate of low-carbon growth since 2012 set California back by at least five years in its quest for a low-carbon future. Diablo Canyon, California’s one remaining nuclear plant, is now scheduled for shutdown in 2025.
Figure 7: Percent average annual generation by source, April 20, 2010 to March 9, 2016. Note that 2010 and 2016 do not contain a full year of data.
Hourly data plots:
Plots of hourly data are messy at any scale exceeding a few months, so here I concentrate on the high-demand period during August and September 2015. Figure 8 plots the hourly data for this period:
Figure 8: Hourly generation by source, August and September,2015
Peak demand of 47.3GW occurred at 5pm on September 10. (Note this is an hourly average; the short-term peak demand would have been higher). Load-following over the two-month period shown was performed dominantly by thermal (gas-fired) plants with an assist from hydro. Renewables and imports did not contribute significantly. The R squared values obtained by correlating individual generation sources with total generation are:
- Thermal 0.83
- Hydro 0.74
- Renewables 0.19
- Imports 0.03
Renewables show a small positive correlation with total generation because the pre-noon increase in solar generation happens to coincide with the morning increase in demand, although the post-noon solar generation decrease does not. Baseload nuclear is of course uncorrelated with total generation.
Figure 9, which shows generation by source during the seven days between September 6 and 12, provides more information on California’s load-following procedures during high-demand periods. Most of the load-following requirements were handled by thermal, but in this case renewables, hydro and imports also contributed :
Figure 9: Contributions of different generation sources to load-following, hourly data, September 6 to 12, 2015
The table below quantifies the load-following contribution from each of the main generation sources between minimum demand and peak demand on September 10 and between peak demand on September 10 and minimum demand on September 11:
About 60% of the load-following requirement was met by thermal while the remaining 40% was met in a roughly even three-way split between hydro, imports and renewables. However, only hydro and imports were cycled in a controlled manner (note how Figure 9 shows imports being cut back during early morning low-demand periods). Renewables made a net overall contribution to load following only because solar generation is zero during early morning low demand periods and positive during the day. The fact that the solar peak occurs several hours before the demand peak creates a different set of concerns that have become known as the California Duck Curve, which is discussed below.
The “California Duck Curve”
Here it is plotted up. The reasons for calling it the “duck curve” are apparent.
Figure 10: The California “Duck Curve”
Figure 10 shows the demand curves that would have to be tracked by load-following generation as increasing amounts of solar generation are admitted to the grid. The curve gets progressively steeper in the late afternoon as more solar generation is added, raising concerns as to whether load-following plants will be able to ramp up quickly enough to keep pace with demand. The problem is thought to most serious on or around March 31, so I plotted up the data for the three day period from March 30 through April 1, 2015. Figure 11 shows total generation (again assumed to be equal to demand) over this period. Ramp rates are generally low on the upside of the demand curve and higher but not excessive on the downside:
Figure 11: Total generation, hourly data, March 30 through April 1, 2015
Figure 12 now superimposes solar generation and shows the resulting demand curve that other generation sources had to track. Ramp rates were about the same on the up and down sides of the demand curve, but nowhere excessive:
Figure 12: Total generation, hourly data, March 30 through April 1, 2015 with solar generation superimposed. The gray bars show the resulting demand curve.
Figure 13 shows how the main generation sources combined to match the Figure 12 demand curve. Most of the load-following was handled by thermal generation, with an assist from hydro and occasional minor contributions from imports and “other renewables” (renewables less solar).
Figure 13: Contributions of different generation sources to load-following, hourly data, March 30 through April 1, 2015
In summary, 2015 seems to have passed without a duck curve problem. But things will deteriorate rapidly if the amount of solar admitted to the grid continues to increase in coming years. Had there, for example, been twice as much solar in 2015 the impact would have been as shown in Figure 15. Ramp rates on the upside of the demand curve would have reached 15GW in three hours, and CAISO is concerned that California’s “very old” steam turbine plants may not be able to handle long-term ramp rates this high:
Figure 14: Total generation, hourly data, March 30 through April 1, 2015 assuming twice as much solar generation. The gray bars show the resulting demand curve.
A final uncertainty remains to be discussed. Do the CAISO data include unmetered solar from rooftop and other private solar installations? If not the CAISO estimates of solar generation will be too low and the duck curve problem worse than shown. According to this comment from Peter Gleick unmetered generation is included in the EIA state grid data but not the CAISO grid data. I took a quick look at the EIA data but was unable to reconcile them with the CAISO data (2014 nuclear and wind generation compare within a percent or two but CAISO shows 14% more solar generation, 16% more total generation and 36% less geothermal generation than EIA). These discrepancies need to be looked into, but that will have to be the subject of another post.