V. PG&E's Cost-Effectiveness Model

In this proceeding, PG&E conducted its cost-effectiveness modeling using a Monte Carlo simulation model. The model has over 100 input variables. Each variable has a low, base and high value associated with it. The low value has a 10% probability of the actual value being below it. The base value has a 50% probability of the actual value being below it. The high value has a 90% probability of the actual value being below it. The variables and their probabilities were estimated by PG&E or its consultants. For each run, the model randomly picks a low, base, or high value for each variable, and compares the cost of the SGRP to the no project alternative. The model then performs successive runs until the mean simulation error was below $10 million.⁷ For PG&E's cost-effectiveness analysis, the model performed over 9600 runs. The mean of the results of the model runs was then calculated to determine the mean net present value of revenue requirements (NPV).⁸

PG&E's analysis yields an NPV of the SGRP of $1.2 billion. This means that there is an 80% probability that the NPV will be between $173 million and $2.44 billion, with a 10% probability that the NPV will be below $173 million and a 10% probability that it will be above $2.44 billion. It also means that there is a 95% probability that the SGRP will be cost-effective, and a 5% probability that it will not be cost-effective.

San Luis Obispo Mothers for Peace, Greenpeace, Sierra Club, Public Citizen, and Environment California (collectively, MFP) contend that PG&E's model has never been relied upon by a regulatory agency for decision making. MFP also states that PG&E's model ignores correlations between variables such as between capital costs, O&M costs, and capacity factor.

As to correlations between variables, it is reasonable to assume that capital costs and O&M costs are related to Diablo's performance as measured by its capacity factors. This is because capital and O&M costs are incurred in order to keep Diablo in operation. It would seem reasonable, for example, to assume that a low capacity factor or an outage could result in increased capital or O&M expenditures to correct any plant problems that led to the low capacity factor or outage. In this case, a low capacity factor would be associated with increased capital or O&M expenditures. However, increased capital additions or O&M could be implemented to avoid a decrease in the capacity factor or an outage, in which case increased capital additions or O&M would be associated with no change in capacity factor. In addition, there are other factors which may influence capacity factors such as regulatory requirements and plant design. For these reasons, we do not find it unreasonable that the model fails to incorporate a mathematical formula directly linking capital costs, O&M costs and capacity factors. We also note that MFP has not indicated what the mathematical relationships between these or other variables should be.

PG&E provided an explanation of its model, and the assumptions it used. As discussed later in this decision, ORA and TURN's models yield results generally similar PG&E's model when similar inputs are used. This tends to indicate that PG&E's model's results are not unreasonable. Therefore, we conclude that PG&E's Monte Carlo simulation model is appropriate for use in this proceeding. We will now address the parties' concerns regarding various model inputs.