VIII. Modeling of DA Cost Responsibility Surcharges for DWR Costs

A. Role of Modeling in Analyzing DA Cost Shifting

As a framework for analyzing DA cost shifting effects, computer modeling simulations were offered into evidence. An initial series of model simulations was performed by DWR through its independent consultant, Navigant, using the PROSYM model. The Commission's Energy Division conducted a technical workshop on April 12, 2002 in which parties agreed upon various modeling scenarios to be performed by DWR. DWR submitted original modeling runs to parties on March 8, 2002, and revised on March 19, 2002, incorporating a base case with 10 scenarios and two sensitivity cases.

DWR's modeling analysis sought to compute DA cost responsibility charges at the level necessary to keep bundled customers' retail rates from increasing to cover any added cost burden caused by customers switching from bundled to DA service between July 1 and September 20, 2001. DWR thus calculates the cost shifting that results from the increase in DA participation between, and including, these dates. DWR computed a levelized fixed charge covering the period from inception of DWR purchases in January 2001 and extending over the next 15 to 20 years, to capture the net change in DWR power costs over the life of its long-term power contracts. The costs for electric purchases for the period from January 17, 2001 through September 20, 2001 were higher than the revenues that the DWR collected from the IOUs. In its calculations, DWR assumed that a pro rata share of the shortfall for this period is covered by DA customers.

A number of parties criticize the Navigant modeling approach. On a policy level, some parties question the merits of relying on long-term modeling at all to determine DA cost responsibility, arguing that any attempts to litigate long-term models will be fraught with controversy, speculative, and an unproductive use of time and resources. Other parties support the use of models to perform long-term forecasts, in general, but take issue with Navigant's modeling, in particular. These parties claim that the Navigant analysis systematically overstates DA customer responsibility for DWR procurement costs. The problems raised are both conceptual (e.g., the focus solely of DWR portfolio costs) and factual (e.g., levels of DA load, market prices, and assumptions regarding sales below prevailing spot prices.)

PG&E, SCE, SDG&E, and ORA all base their DA CRS calculations on Navigant Scenario 8. Certain parties representing DA interests, however, propose that instead of the Navigant model, the Commission rely on an alternative model sponsored by Henwood Energy Services, Inc. (Henwood) as the basis for forecasting DA CRS. The Henwood Model was offered into evidence through the testimony of J. Richard Laukhart, Director of Henwood (Exh. 31). Henwood presented modeling results on behalf of a consortium of parties, performing a quantitative analysis of the impact of the increase in DA customers on DWR costs and to review Navigant's work.

Henwood modeled a "base case" that represented a revision of Navigant's original base case, updated to reflect Henwood's assumptions. Henwood estimates that the indifference costs associated with DWR power over the years 2002 through 2011 would be $1.96 billion higher than determined under the Navigant modeling, resulting in a lower DA CRS.²⁶

We address the merits of the differences between the Navigant and Henwood modeling in Section XIII below. The different DWR cost values under the models are summarized below:

B. Summary of Proposed CRS Based on Modeled Results

An overall summary of parties' proposed DA CRS cost elements is set forth in Appendix A. We summarize parties' proposals below, regarding DWR costs. Proposals for the URG component are discussed in Section XIV.B.

PG&E

PG&E's proposal for calculating DWR-related DA CRS is based on the Navigant approach Scenario 8, which results in a DWR charge of 3.277 cents/kWh for 2003, declining to 1.878 cents/kWh by 2007.²⁷ PG&E proposes that the Commission direct DWR to present final calculations consistent with its proposed revenue requirement in the DWR revenue requirement proceeding, and that the DWR DA CRS be adopted annually in that proceeding.

SCE

SCE is also generally supportive of the Navigant Scenario 8 modeling approach. SCE indicates that using the Navigant approach, it would require a DA CRS of 2.6 cents/kWh to recover the applicable share of the 2003 DWR revenue requirement.

SDG&E

SDG&E proposes that its DA CRS be set based on an initial 15-year statewide levelized annual cost of 1.22 cents/kWh, subject to correction of DA load figures as specified in its testimony.²⁸ Based upon Navigant Model Scenario 8, SDG&E's utility-specific 15-year levelized DA CRS would be 2.76 cents/kWh. Full recovery of the DA CRS for SDG&E's share of the 2003 DWR revenue requirement would be 4.48 cents/kWh.²⁹ SDG&E's DA CRS is higher than the other utilities because SDG&E has a higher percentage of load served by DWR without a proportionately higher DA load over which to spread the costs.

ORA

ORA's 2003 DWR DA CRS charges for forward DWR costs is $42.52/MWh, and its proposed historical DWR charge is $11.95/MWh.

CLECA

CLECA proposes that the DWR costs be levelized over a period of 10-15 years. Applying its total portfolio approach under the Henwood Base Case, CLECA calculates a DA CRS of $21.69/MWh for 2002 declining to $7.02/MWh by 2011. Appendix C shows CLECA's calculation of actual annual forward costs for the period fourth quarter 2001 through 2011 based on two Henwood model scenarios. In each scenario, the total portfolio DA CRS-related costs drop significantly over the 10-year period. The DA CRS in the earlier years under the total portfolio method are lower than the DWR-only CRS, reflecting the effect of the lower-cost URG power on the bundled portfolios.

CLECA calculates levelized forward DA CRS using a 15-year recovery period, both with an initial implementation at the start of the fourth quarter of 2002 and with a two-year delay under which utility undercollections would be collected from DA customers first. The 15-year levelized charges with a two-year delay and associated financing costs, fall at or below $14.25/MWh. Without the two-year delay, the Henwood scenarios would fall at or below $13.52/MWh.

CLECA also presents calculations of its proposed 15-year levelized cost on a utility-specific basis. Under the Henwood Base Case on Table 1, the statewide 15-year levelized cost, with a two-year delay, is $14.25/MWh. The comparable figure on Table 2 for PG&E is only $7.51/MWh, while the figures for Edison and SDG&E are $19.17/MWh and $29.08/MWh respectively. Comparable figures for the three utilities using the Henwood Base Case with higher Market Prices are $5.11/MWh, $15.93/MWh and $25.48/MWh.

The levelized forward costs for PG&E under the Henwood Base Case Scenario are $7.51/MWh over 15 years with a two-year deferral in recovery, or $7.12/MWh with no deferral. These figures exclude any bond charge. In order to mitigate the effects of higher charges applicable to SDG&E, CLECA recommends a 20-year recovery period.

CIU

CIU proposes the application of a 20-year levelized charge on a statewide uniform basis. Thus, CIU does not calcuate separate utility-specific charges. CIU calculates a levelized annual charge of 1.225 cents/kWh over a 20-year period covering both an historical and an ongoing DWR component. CIU calculates nonlevelized ongoing charges starting in 2002 of 3.00 cents/kWh declining to 1.027 cents/kWh by 2010, and terminating thereafter. CIU relies on Henwood's modeling assumptions. CIU assumes that power will be sold off system at 100% of the PROSYM forecasted spot price. (See CIU/Chalfant, Exh. 33, pp. 7-8.)

CMTA

CMTA proposes the use of a 20-year levelized charge that takes into account the total utility portfolio. CMTA's forecasts of the levelized charge for the ongoing portion of costs is 0.407 cents/kWh, covering the DWR contract costs for the 2002-2011 period. CMTA adds a component of 0.285 cents/kWh to amortize the previously incurred DWR costs during the historic period through the fourth quarter of 2001. This calculation is set forth on Table 3 of Exh. 39, Testimony of Beach, as reproduced in Appendix D of this decision.

C. Merits of Multi-Year Modeling Versus Annual Forecasting

1. Parties' Positions

A number of parties call into question the whole rationale for modeling multi-year forecasts as a basis for DA CRS. Whether the modeling is performed by Navigant, Henwood, or another entity, the reliability of long-term forecasts remains in question. Parties are in dispute over whether the Commission should rely at all on multi-year modeling forecasts of DWR costs or should simply perform one-year-ahead forecasts of costs subject to annual true ups.

The modeling efforts performed in this proceeding by DWR/Navigant and Henwood involved running complex models with multiple assumptions to forecast annual revenue requirements through 2011. In order to develop revenue requirements over such a long-time horizon, DWR and Henwood necessarily assembled long-term cost forecasts of total load and net short load on a statewide basis, new capacity additions and gas prices, among other items. For the utility-specific analysis, additional assumptions regarding inter-utility contract allocation are required.

The advantage of performing multi-year forecasts of cost responsibility is that the effects of relatively high costs in the early years can be combined with declining costs in later years to yield a "levelized" annual charge. The levelized charge minimizes the burden on DA customers in the early years by deferring a portion of those costs into later years, with the deferred portion financed at an assumed cost of money.

Several parties express concern, however, regarding the uncertainty surrounding forecast assumptions, particularly as they extend farther into the future. Adopting a method of determining DA charges that requires a long-term forecast will serve to make for more contentious proceedings. A levelized charge methodology results in charges for 2003 being affected by the assumptions made about market conditions as far in the future as 2011. Although long-range forecasts are necessary to evaluate long term-term decisions such as the purchase or construction of capacity, various parties argue that such forecasts are not needed for assigning cost responsibility in this case. Rather than litigating long-term forecasts in this proceeding, these parties propose setting the DWR charge on an annual basis.

In addition to the uncertainty associated with developing a long-term forecast, levelizing costs may create inter-year cost shifts between bundled and DA customers. In DWR's scenarios, the cost per kilowatt-hour (kWh) increase to bundled customers attributable to DA tends to be highest in 2003, and declines through 2011.

TURN raises the concern that levelized fixed charges have the potential effect of forcing bundled service customers to lend DA customers money at an effective 7.1% interest rate with up to a 20-year term. Based on DWR's Revised Base Case, TURN computes that bundled service customers would pay $200 to $300 million extra annually over the 2002-2008 time frame (by as much as $300 million per year in 2002-2004). On a present-value basis, they could pay $1.5 billion more by the end of 2008 if a levelized direct access loan was adopted. The $1.5 billion in excess payments would then be repaid through 2021.

TURN argues that if direct access provides such a benefit to its recipients, it should at least stand on its own feet without subsidized low-interest financing from bundled service customers. If the Commission does implement a levelized charge, TURN believes that a balancing account should be established, ensuring that the financing costs used to benefit direct access customers remain within each customer class.

As an alternative to computing a levelization of future forecasts, various parties propose, instead, simply adopting an annual cap on the maximum amount to be paid by DA customers in any given year. Any excess over the cap would then be deferred into future years.

2. Discussion

We conclude that long-term models serve a useful role in this proceeding, but not for the purposes of setting a levelized annual charge. We decline to rely upon the multi-year modeling forecasts presented by any of the parties in this proceeding as a basis to set specific levelized annual charges applicable to DA customers. We agree that the assumptions made regarding key variables extended several years into the future are too uncertain to form a basis for setting specific levelized charges in this decision. We still find that the multi-year forecasts are relevant, however, by providing more generalized indications of longer-term trends in the relative trend of DA CRS of uneconomic costs over time. Under both the Navigant and Henwood modeling assumptions, we can generally conclude that the magnitude of uneconomic costs are likely to be greater in the initial years and will decline in later years of the DWR contracts. As a result, we conclude that it is feasible to establish caps on the maximum DA CRS amount in the first few years, such that the unrecovered balance could be made up with surplus collections from DA customers in the latter years when uneconomic costs decline. We discuss the capping issue in Section XV.

For purposes of setting the DA CRS effective for the year 2003, we need only to rely on a single year-ahead forecast. We further conclude that there should be consistency between the forecast assumptions underlying the DWR charges paid by bundled customers and by DA customers. Otherwise, the use of inconsistent forecast assumptions would result in either under- or over-recovery of the respective shares of DWR costs from bundled and DA customers, and our goal of bundled customer indifference would be undermined. Since the DWR power charges applicable to bundled customers is being determined in A.00-11-038 et al., we shall require that the assumptions underlying the calculation of DA CRS be consistent with the 2003 DWR/Navigant modeling underlying the revenue requirement implemented in the A.00-11-038 et al. proceeding.

Various parties representing DA interests argue that the modeling work performed by Navigant is unreliable for use in this proceeding, and the modeling performed by Henwood is superior and should be the sole model relied upon for assessing costs and charges in this proceeding. CIU, in particular, notes the series of modeling mistakes and corrections made by Navigant through the course of this proceeding.³⁰ CLECA also claims that the Navigant scenarios suffer from major analytic flaws, including too high an estimate of new power plants, which depresses market clearing prices (MCP) for electricity, and too low an estimate of prices for off-system sales (50% of MCP).

DWR replies that the majority of parties' criticisms of its modeling focus on DWR's "base case" which was first submitted in March 2002, but completely ignore the updates reflected in DWR's Scenarios 1, 7, and 8, distributed to the parties in late April 2002. DWR argues that these updated scenarios responded to the majority of the criticisms raised by parties, and that most of parties' criticisms relate to the now outdated March 2002 version of the model.

We note that, despite their differences, there are a number of similarities between Navigant and Henwood's modeling efforts. Both used the same basic modeling tools: Henwood's "Electric Market Simulation System" and accompanying NERC database. Both also used Henwood's production simulation model, PROSYM. Navigant's starting point was Henwood's publicly released NERC database circa the fall of 2000, which it then modified. Henwood's analysis relied on its most recent NERC database released in the spring of 2002, which Henwood then modified. From that point, the two firms took separate approaches to performing the DA CRS modeling.³¹ Given the common elements between the modelers, we see no necessity to elevate one modeler over the other as having a favored position for future modeling assignments. The participation of Henwood in this proceeding contributed positively to the quality of the modeling and forecasting data in the record. Having multiple modelers offers a broader perspective and a forum for more critical evaluation of modeling conventions. We review the model's results and adopt a prescribed methodology for calculating the DA CRS components for ongoing costs in Section XIII.