Combined cycle (CC) generators are attractive for many reasons, not the least of which is their low heat rate (HR), which is due to the combination of one or more combustion turbines (CT) with one or more heat recovery/steam turbines (CA). To develop some intuition about the advantage provided by the CA part of the CC, we used hourly EPA CEMS data to calculate HR for the CT part of a sample plant in NY ISO. This has importance for predicting fuel use by CC plants, particularly during summer months.

The first chart is a scatter plot comparing HR to capacity factor (CF) for July 2018. Compare this plot to HR for the entire CC plant (see CEMS & Market Heat Rate tab below). It is apparent that the CT heat rate is significantly higher.



The scatter suggests that the CTs are mostly operated at 80 to 90 percent CF. This is confirmed by the histogram, which shows two modes: a lesser mode in the 0.50 range and the main mode in the 0.80 range. This is cleary the sweet spot for the plant.



What is the take away from these two graphs? The CT units consume the fuel. (During July, CFs are high enough to obviate duct firing.) So, calculating fuel use based on the plant HR may under-estimate fuel use depending on the kind of calculation.

Combining EPA's hourly data from continuous emissions monitoring systems (CEMS) with EIA's electric power and fuel use data creates an information source using public data only, thus providing provenance and repeatability. There are two keys to using CEMS as a source of hourly load and heat input data: understanding its taxonomy and access to a crosswalk between EPA generators and their EIA equivalents. We recently obtained a draft crosswalk. But first the EPA taxonomy.

The highest CEMS taxon is the plant, whose identifier is ORISPL. Each plant consists of one or more units identified by CEMS unitId. Units consist of one or more generators identified by generatorId. CEMS unit identifiers are often unique to EPA. (Because facilities data are exposed as a JSON object, plants, units and generators are array elements and therefore also identified by array index.) The CEMS facilities list contains 1,745 plants comprising 5,473 units and 6,985 generators. (Not all of these are reported as operational.)

CEMS collects data from air emissions sources of 25MW or more. It exludes the majority of the 21,438 generators in the EIA-860 file - only 7,102 have a nameplate of 25MW or greater. It also excludes generators that do not combust fuel, thereby omitting the steam part of combined cycle plants/units among others. (See CEMS vs. EIA Data: Example tab below.)

The draft crosswalk covers 1,721 plants - slightly fewer than contained in the facilities file. There are 6,861 unique crosswalk entries i.e. EIA data rows that have a CEMS counterpart. So, it is possible to use CEMS data in a mashup that adds hourly data to EIA's already rich data store.

Hourly CEMS data were mashed with EIA-860 and -923 to produce a scatter plot of operating heat rate vs. capacity factor for February and July, 2018. The plant was Bethlehem Energy Center (see previous tab). Results show that the plant will run and cover its fuel cost when the market heat rate is above roughly 8,000 Btu/kWh. As the market rate falls, the plant can continue to cover fuel expense if it can operate at a higher capacity factor and therefore a lower heat rate.



Combustion (CT) units were combined with the heat recovery (CA) to derive hourly capacity factors and heat rates. Although minor amounts of gas were fired in the duct burner during February, this was ignored.

At least three tranches are evident in the plot: capacity factors below 40% or so, capacity factors between 35% and 50% with heat rates between 7,000 and 8,000, and the main tranche above 50% that contains most of the points. Points in the main tranche follow a well developed trend and likely represent the plant's primary operating doctrine. Fuel use predictions based on market heat rate and demand within the ISO are likely more reliable in this tranche.

How does EPA's continuous emissions monitoring system (CEMS) data compare with EIA's generation and fuel use data? Only a comprehensive comparison might provide a really good answer but a quick comparison involving a randomly chosen NY ISO plant provides some intuition. February and July, 2018 were chosen to include seasonal differences. Conclusion? Relying on CEMS for heat rates and fuel use may require substantial care.

The example plant is Bethlehem Energy Center (ORISPL 2539), Albany, NY which came online in 2005. This is a gas-fired combined cycle plant consisting of three combustion turbines (CTs) of 194.3 MW nameplate each which feed exhaust into a header to a heat recovery unit for one 310.2 MW steam turbine (CA). There is duct firing but no CA bypass.

EIA-923 tabulates monthly combined generation (MWh) and heat input (MMBtu) for the CTs and for the CA. CEMS tabulates hourly load (MW) and heat input (MMBtu) for each of the CTs but no data were found for the CA. The following table shows that EIA and CEMS nearly agree as to the CTs. (CEMS unit IDs are used for the CTs.)

Differences are minor and due in part to how data are collected, reported, and used here. Notably, CA data are not included in CEMS.

If CEMS data are used to calculate heat rates for use in downstream application like stacking models or to predict fuel consumption, omitting CA data could be important. The following table shows calculated capacity factors and heat rates.

Capacity factors are nearly identical in CEMS and EIA and CT heat rates are very close. Capacity factors suggest that the CA unit is dispatched more or less like the CTs. However, CEMS does not include the heat recovery, CA, unit so plant heat rate is much higher for CEMS. The difference may be material for downstream applications of the data.

Scatter plots show how combined cycle (CC) generating plants were dispatched in the three ISOs comprising the north-east section of the country during January - July, 2018. Each point represents the plant-level capacity factor (CF) calculated from the combustion turbine(s) (CT) and the heat recovery (CA) units at a single plant during a single month.

The 45 degree reference line is the locus of points at which CT and CA CF are equal i.e. their respective generation (MWH) divided by the product of monthly hours (H) and nameplate capacity (MW) are equal. Because CAs recover waste heat from CT exhaust, their dispatch is important for overall CC heat rate and therefore fuel consumption as well as dispatch modelling. Interestingly, these plots suggest that CA generators are not dispatched like their associated CTs.





Data were developed by mashing EIA-860 and EIA-923 and discarding questionable points e.g. CFs greater than 1.00. Further analysis is on-going.

The following plot shows average load for each hour of the day for NY ISO during summer and winter months. Loads were normalized by dividing each hour's average load by minimum load. Normalization allows comparison without changing the shape of each curve. An afternoon ramp does not appear to occur during summer months for the entire ISO. During winter months load tends to fall around noon to a local minimum and then rise again until the evening hours.



LCs were also analyzed by NY ISO zone to look at geographic differences. Three zones - see the map below - are compared with the ISO as a whole for July - September 2018. (Zonal demand only just became available from EIA for some ISOs beginning in July.) The take-away from this plot is that an afternoon ramp may be present in some zones and not others or in a larger geographic area. Further analysis of this issue is on-going.

The following plots show 2018 average hourly native load for ERCOT and for three of its eight control areas. Three months are shown in order to illustrate differences by time of year and geography. The ERCOT native load report was used for this study. Each month's hourly load was normalized to the month's mimimum in order to facilitate comparison. A map of ERCOT control zones appears at the bottom of this tab.

The plots show that an afternoon ramp is most likely to happen during winter/cooler months. At other times, native loads increase through beginning sometime between 4:00 and 6:00 AM and peaking in early evening.

Loads on ERCOT generators vary through the day in response to demand, which includes both native load and exports. Weather, which depends in part on season/month, is a major determinant of demand. The following plot shows how the average daily load varies by month. Each daily curve is normalized to the curve's minimum to facilitate comparison.



The plot shows daily curves changing shape over the course of 2017. This channge can be measured by means of a load curve metric which appears to remain relatively stable from year to year because the shape of the curve does not vary much. The following chart shows the metric for each month.



More comparisons and more information about the metric will be published over the next week.