California Renewable Portfolio Standard: Renewable Generation Integration Cost Analysis
Multi-Year Analysis Results and Recommendations

Publication Number: CEC-500-2006-024
Publication Date: month 2006
PIER Program Area: Renewable Energy Research

The executive summary, abstract and table of contents for this report are available below. This publication is available as an Adobe Acrobat Portable Document Format File. In order to download, read and print PDF files, you will need a copy of the free Acrobat Reader software installed in and configured for your computer. The software can be downloaded from Adobe Systems Incorporated's website.

---

Executive Summary

The California Renewables Portfolio Standard^[1] requires a "least-cost, best-fit" strategy for selecting new generation projects to fulfill its renewable energy supply goals. This explicitly includes indirect integration costs in the bid evaluation process. In previous work^[2,3], integration costs were identified, valuation methodologies were defined, and a one year analysis of 2002 was performed.

This report documents a multi-year analysis of integration costs, applying the previously defined methodologies to a three year period from 2002 to 2004. The multi-year analysis provided opportunities to verify the consistency of the methodologies, further examine the practical issues associated with integration cost analysis, and to study the impact of renewables on integration costs over several years.

The methodologies are straightforward and were applied with little modification from their implementation in the previous one year analysis; the changes that were made are documented herein. The input data required for the analysis, however, was more problematic. Data quality and confidentiality issues hindered the progress of the study. The most critical data issues were ultimately resolved by using a combination of datasets from CaISO, SCE, and PG&E; performing extensive manual reviews of the data using custom developed programs; and training personnel who had access to the data to perform the analyses. However, outstanding data issues limited the analysis, as detailed within the report.

Overall, the multi-year integration cost analysis results were reasonable, consistent with the analysis results of the previous one year dataset and, in some cases, verified with alternate approaches. The results of the capacity credit analysis are summarized in the table below.

The capacity credit analysis uses a conventional medium gas unit as a benchmark. Because of inconsistencies in the nameplate capacities provided for the generation aggregates, results are presented relative to both reported nameplate capacity and annual peak generation.

Biomass has outage rates comparable to the gas benchmark unit and, therefore, a high capacity credit. The geothermal outage rates are lower than the benchmark unit, resulting in a capacity credit exceeding 100%. The solar values are relatively high, as expected given its natural tendency to track load and the plants' auxiliary gas generators. Wind values ranged from 27% to 44% (based on annual peak generation; 24% to 39% based on reported nameplate capacity), with both regional and inter-annual variation. This is reasonable given wind's variable nature. The results were verified using an alternate method.

The results of the regulation analysis are summarized in the table below. Negative values indicate a cost.

The resources studied have fairly minor impacts on total system regulation requirements. There is some inter-annual variation; in most cases, the changes follow the cost trend of actual regulation commitment by CaISO between 2002 and 2004. Because of the sheer size of total load, its regulation cost is consistently very close to that of the total system requirement. Geothermal, with a fairly flat output, has a low regulation cost, but a slightly higher value in 2002 when it was block scheduled for part of the year. The regulation costs of the solar and wind aggregates range between $0.24/MWh and $0.70/MWh, ignoring the anomalously low value for wind in San Gorgonio in 2002. While these values are higher than the results for biomass and geothermal, they are still quite modest. The solar results are consistent with the minute-to-minute variability in its generation data. The regulation costs imposed by wind are reasonable given that there are no apparent mechanisms that tie wind plant performance to the power system's needs either favorably or unfavorably in the regulation time frame.

The results of the load following analysis are summarized in the table below.

Error	2002				2003				2004
	AVERAGE MINIMUM		AVERAGE MAXIMUM		AVERAGE MINIMUM		AVERAGE MAXIMUM		AVERAGE MINIMUM		AVERAGE MAXIMUM
	MW	Compared to load forecast error alone (%)	MW	Compared to load forecast error alone (%)	MW	Compared to load forecast error alone (%)	MW	Compared to load forecast error alone (%)	MW	Compared to load forecast error alone (%)	MW	Compared to load forecast error alone (%)
Load forecast alone	-1945	100%	2112	100%	-1600	100%	2151	100%	-1439	100%	1529	100%
Load scheduling alone	-4747	244%	1302	62%	-4021	251%	2158	100%	-3700	257%	1776	116%
Scheduling bias	-5337	274%	1708	81%	-3336	208%	1534	71%	-3016	210%	1634	107%
Combined load forecast and renewable resource scheduling error
Biomass	-1944	100%	2115	100%	-1603	100%	2157	100%	-1432	100%	1536	100%
Geothermal	-1947	100%	2112	100%	-1599	100%	2149	100%	-1442	100%	1529	100%
Solar	-1897	98%	2055	97%	-1631	102%	2153	100%	-1467	102%	1541	101%
Wind (Northern Cal)	-1946	100%	2148	102%	-1591	99%	2203	102%	-1419	99%	1554	102%
Wind (San Gorgonio)	-1930	99%	2142	101%	-1581	99%	2163	101%	-1443	100%	1545	101%
Wind (Tehachapi)	-1931	99%	2177	103%	-1569	98%	2181	101%	-1435	100%	1544	101%

The combined load forecast and renewable resource scheduling error values above indicate that renewables do not have a significant effect on the total energy requirements from the short term load following market at current penetration levels. The minimum scheduling bias was well over 200% greater than the combined forecast and scheduling error, implying that ample depth is available in the short term generator stack to handle incremental energy requirements.

A complementary methodology for analyzing ramping capability and requirements is also presented with a preliminary analysis. The ramping capability of thermal generators responding in the load following time frame appears to very large and capable of supporting a large amount of renewables. The ramping requirements of intermittent renewables appear to be significantly lower than the requirement of the total system load and the capability available in the CaISO control area.

Provided the necessary data with sufficient quality, integration cost analysis becomes a relatively quick and straightforward process. An Integration Cost Analyst (ICA) is proposed to perform and report on integration cost analysis on a regular basis. It is recommended that the California Energy Commission or CPUC dedicate personnel and resources to perform the functions of an ICA. However, given the data issues encountered during this study, the tasks of handling/preparing data and analyzing integration costs should be made distinct and separate. This would also benefit other recent and current studies which require similar data. A data handling entity is proposed who would coordinate with data sources (CaISO, IOUs, etc.) and the ICA to ensure the availability of good data quality as needed.

FOOTNOTES

1. Sher et al, California Senate Bill 1078; Chapter 516, Statutes 2002; chaptered September 12, 2002.


2. Kirby, B., M. Milligan, Y. Makarov, D. Hawkins, K. Jackson, and H. Shiu, "California RPS Integration Cost Analysis – Phase I: One Year Analysis of Existing Resources", Consultant Report 500-03-108C, California, California Energy Commission, December 2003.


3. Kirby, B., M. Milligan, Y. Makarov, D. Hawkins, J. Lovekin, K. Jackson, and H. Shiu, "California Renewables Portfolio Standard Renewable Generation Integration Cost Analysis – Phase III: Recommendations for Implementation", Consultant Report P500-04-054, California, California Energy Commission, July 2004.

---

Abstract

None available.

---

Table of Contents

Table of Contents i

List of Figures iii

List of Tables v

Abbreviations vi

Nomenclature viii

Acknowledgements x

Summary xi

1 Introduction 1

2 Capacity Credit 4

3 Regulation 25

4 Load Following 43

5 Data 67

6 Recommendations 92

References 95

Appendix A: Control Performance Standards 96

Appendix B: Regulation Allocation Methodology 98


LIST OF FIGURES

Figure 1.1 How integration costs fit in the least-cost, best-fit process. 2

Figure 2.1. Hourly LOLP, ranked, 2002-2004. 11

Figure 2.2. Load in 2002 during top risk hours, ranked by load, LOLP, and load net hydro and interchange. 12

Figure 2.3. LOLP in 2002 at top hours of load net hydro and interchange. 13

Figure 2.4. Load in 2003 during top risk hours, ranked by load, LOLP, and load net hydro and interchange. 14

Figure 2.5. LOLP in 2003 at top hours of load net hydro and interchange. 14

Figure 2.6. Load in 2004 during top risk hours, ranked by load, LOLP, and load net hydro and interchange. 15

Figure 2.7. LOLP in 2004 at top hours of load net hydro and interchange. 15

Figure 3.1 Decomposition of hypothetical weekday morning load. 26

Figure 4.1. Forecast and actual load over a three day sample period. 49

Figure 4.2. Scheduled and actual load over a three day sample period. 50

Figure 4.3. Actual and scheduled wind generation over a three day sample period. A simple persistence model was used to produce the schedule. 51

Figure 4.4. Thermal ramping capability and load ramping requirement. The 2002 and 2003 load ramping requirement traces (purple and red) are nearly overlapped. Thermal ramping capabilities typically exceed load ramping requirements. 56

Figure 4.5. The ratio of simultaneous load ramping requirements and thermal generation ramping capability, 2002. Thermal ramping capability exceeds load ramp requirements more than 97% of the time. 57

Figure 4.6. The total wind ramping requirement in California, 2002, calculated from ten minute averages. 58

Figure 4.7. The total wind ramping requirement in May 2002, showing large ramp-up requirements. 59

Figure 4.8. The total wind ramping requirement in February 2002, showing large ramp-down requirements. 60

Figure 4.9. Typical total wind ramping requirements, shown in July 2002. 61

Figure 4.10. Typical total wind ramping requirements, shown in September 2002. 61

Figure 4.11. Solar ramping requirement in California, 2002, calculated from ten minute averages. 62

Figure 4.12. Solar ramping requirements, June 2002. 62

Figure 4.13. Solar ramping requirements in July 2002, showing large ramp-down requirements. 63

Figure 4.14. Solar ramping requirements in September 2002, showing a large ramp-up requirement. 63

Figure 4.15. Comparison of wind ramping requirements calculated with 10 minute and hourly data. 64

Figure 4.16. Ramping requirements for wind and solar aggregates based on hourly data; the requirements are small compared to the load ramping requirement. 64

Figure 4.17. The ramping requirement of wind in Tehachapi based on hourly data over three years. 65

Figure 4.18. The ramping requirement of solar based on hourly data over three years. 65

Figure 5.1. Generation of the biomass aggregate in the CaISO multi-year dataset. Two years from fall of 2002 to fall of 2004. 73

Figure 5.2. Generation of the biomass aggregate in the CaISO multi-year dataset. One week from winter of 2004. 73

Figure 5.3. Generation of the SCE territory geothermal aggregate in the CaISO multi-year dataset. Two weeks in winter of 2002. 74

Figure 5.4. Generation of the SCE territory geothermal aggregate in the CaISO multi-year dataset. One month in spring of 2003. 74

Figure 5.5. Generation of the solar aggregate in the CaISO multi-year dataset. One year from summer of 2003 to summer of 2004. 75

Figure 5.6. Generation of the solar aggregate in the CaISO multi-year dataset. One month in summer of 2004. 75

Figure 5.7. Generation of the Northern California (Altamont, Pacheco, Solano) wind aggregate in the CaISO multi-year dataset. January 2002 to September 2004. 76

Figure 5.8. Generation of the Northern California (Altamont, Pacheco, Solano) wind aggregate in the CaISO multi-year dataset. One month in summer 2004. 76

Figure 5.9. Generation of the San Gorgonio wind aggregate in the CaISO multi-year dataset. January 2002 to September 2004. 76

Figure 5.10. Generation of the San Gorgonio wind aggregate in the CaISO multi-year dataset. One month in summer of 2003. 77

Figure 5.11. Generation of the Tehachapi wind aggregate in the CaISO multi-year dataset. January 2002 to September 2004. 77

Figure 5.12. Generation of the Tehachapi wind aggregate in the CaISO multi-year dataset. One month in summer of 2003. 77

Figure 5.13. One day from the CaISO multi-year dataset showing a large dropout. 83

Figure 5.14. Three days from the CaISO multi-year dataset showing a data spike. 83

Figure 5.15. A twelve hour period from the CaISO multi-year dataset showing a small dropout. 84

Figure 5.16. One day from the CaISO multi-year dataset showing a sharp 40 MW drop suspected to be a partial dropout in the data aggregate. 84

Figure 5.17. One year of data showing artificially elevated data floors. The red trace is from the CaISO multi-year dataset. The light blue trace behind it is corresponding IOU data. 85

Figure 5.18. Almost three years of data showing artificially elevated data floors. The red trace is from the CaISO multi-year dataset. The light blue trace behind it is corresponding IOU data. 85

Figure 5.19. Three years from the CaISO multi-year dataset showing an occurrence of the dropout/interpolation error. 86

Figure 5.20. One week from the CaISO multi-year dataset showing a dropout/interpolation error immediately followed by a data spike. 86

Figure 5.21. Screenshot of one of the programs developed to process the CaISO multi-year dataset using IOU data as a basis of comparison. 89

Figure 5.22. Screenshot of one of the programs developed to process the CaISO multi-year dataset using IOU data as a basis of comparison. 90

Figure B.1. The relationships among the regulation components (A and B) and the total if A and B are positively correlated (top), negatively correlated (middle), or uncorrelated (bottom). 99

Figure B.2. The relationship among the regulation impacts of loads A and B and the total (T) when A and B are neither perfectly correlated nor perfectly uncorrelated. 100

Figure B.3 Application of the vector-allocation method to the case with more than two loads. 101


LIST OF TABLES

Table 2.1. California peak demand hours for four years from 2001-2004. Times are in Pacific Standard Time. 5

Table 2.2. Step-by-step description of ELCC based capacity credit analysis methodology. 8

Table 2.3. Capacity credit analysis results, based on annual peak generation. 17

Table 2.4. Capacity credit analysis results, based on rated capacity reported by the IOUs. 18

Table 2.5. Capacity credit results with hydro and interchange removed; results based on annual peak generation. 19

Table 2.6. Capacity credit results with hydro and interchange removed; results based on rated capacity reported by the IOUs. 20

Table 2.7. Capacity factor over peak hours based on annual peak generation. 20

Table 2.8. Capacity factor over peak hours based on rated capacity reported by the IOUs. 21

Table 2.9. ELCC compared to peak capacity factors (June through September, weekdays excluding holidays, 12:00 p.m. to 6:00 p.m.) for three years, based on rated capacity reported by the IOUs. 22

Table 2.10. ELCC with hydro and interchange excluded compared to peak capacity factors (June through September, weekdays excluding holidays, 12:00 p.m. to 6:00 p.m.) for three years, based on rated capacity reported by the IOUs. 23

Table 3.1 Regulation inputs/outputs: Verify data consistency. 31

Table 3.2. Regulation inputs/outputs: Calculate total system compensation requirement. 31

Table 3.3. Regulation inputs/outputs: Estimate short term forecast with 15 minute rolling average. 32

Table 3.4. Regulation inputs/outputs: Calculate raw regulation component by subtracting short term forecast. 33

Table 3.5. Regulation inputs/outputs: Calculate total system regulation less resource of interest. 34

Table 3.6. Regulation inputs/outputs: Calculate statistical metrics. 34

Table 3.7. Regulation inputs/outputs: Allocate regulation share for each generator. 35

Table 3.8. Regulation inputs/outputs: Calculate actual regulation share for each generator type. 36

Table 3.9. Regulation inputs/outputs: Calculate actual regulation cost for each generator type. 36

Table 3.10. Original and corrected results of the Phase I (one year, 2002) regulation analysis. Negative values are costs to the system. 38

Table 3.11. Results of regulation analysis of multi-year dataset. Negative values are a cost. 39

Table 3.12. Actual regulation amounts committed in the CaISO control area, 2002-2004. 40

Table 4.1. Load following inputs/outputs: Calculate load forecasting error. 46

Table 4.2. Load following inputs/outputs: Calculate system scheduling error. 46

Table 4.3. Calculate system scheduling bias. 47

Table 4.4. Load following inputs/outputs: Calculate hour ahead schedule of generation resources. 47

Table 4.5. Load following inputs/outputs: Calculate the resource scheduling error. 48

Table 4.6. Results of multi-year analysis of forecast and scheduling errors during peak hours. 52

Table 4.7. Power requirements and generation mix of CaISO in 2002. Data from Platts BaseCase. 53

Table 4.8. Thermal generator ramping capabilities, CaISO in 2002. 55

Table 5.1. Minimum input data requirements for integration cost analysis. 67

Table 5.2. OASIS data used in the multi-year analysis. 68

Table 5.3. CaISO one year 2002 dataset. 69

Table 5.4. CaISO multi-year dataset. 71

Table 5.5. SCE dataset. 78

Table 6.6. PG&E dataset. 79

Table 5.7. Comparison of reported nameplate capacities and annual peak generation of selected generation aggregates from the PG&E and SCE datasets. 82