The models that underlie these valuations reflect over ten years of evolving analysis. Central to this analysis has been developing a full understanding of how fuel costs and weather drive power prices. At the core, RisQuant models OnPeak power prices as a complex spark spread, modulated by weather. We develop a valuation for a particular contract by simulating with daily (1x16) granularity the full spectrum of weather and fuel price scenarios for that term. We have adapted these simulations to produce the other measures that we report.

The RisQuant valuation, then, is a risk neutral, fair value, estimate of where  a conventional OnPeak forward contract will finally settle.

• This valuation is conditional on the current forward fuel curves.
• Delivered volatility is the standard deviation of those simulated outcomes, assuming that current fuel conditions prevail when delivery commences. For the most part, it reflects the influence of historical weather patterns.
• Forward volatility and delta sensitivity of power with respect to natgas both reflect the expected response of power prices to changes in fuel conditions given the current forward natgas volatility curve. It characterizes price movement prior to delivery.

The principal advantage of the RisQuant approach is that we are able to model reasonable, independent estimates for value, volatility and power/fuel price sensitivity for regions that have limited trading. These estimates are easily incorporated into front, middle and back office systems for a variety of purposes. Moreover, we have subjected our models to rigorous validation procedures.

The principal disadvantage is that it is not a true market tested quotation. It differs from a true forward price in two respects. First, the RisQuant value is not derived from trading. Rather, the values are derived from an analysis of prior outcomes to model future outcomes. Moreover, remember that forward power markets have not been risk neutral; they have exhibited (mostly) short aversion tendencies.

Secondly, the RisQuant approach does not take into account expected changes in the mix of generation resources, transmission upgrades and changes in load patterns. However, we have learned that power prices are not particularly sensitive to these factors in the time scales we report, one to thirty-six months. Even while fuel prices and weather have been chaotic, we have learned that the essential underlying structures and relationships have remained remarkably consistent over the years.

• Survey methods comply most closely with the preferences of FAS 133 and FERC. However, only a few markets are liquid enough to yield reliable values. Only ICE 10-X complies fully with FERC price index mandates, and that only for a very few markets. Further, there are even fewer reliable sources for correlations and volatilities that rely on price quotations. The power markets are many years away from comprehensive true forward curves and true transparency.

• Nodal models are the most popular. Essentially, they anticipate the results of the ISO auction process by mimicing it. They model the entire transmission network, along with generation and load forecasts and solve for optimality. Obviously, the computational issues are daunting. Nevertheless, nodal models are the only effective means to evaluate the impact of new generation and transmission.

• Neural network models use sophisticated data mining methods to determine which factors and patterns are currently most important and then use that information to project forward.

• Statistical models rely on static regression models. They typically incorporate mean reversion, persistence, jump diffusion and regime switching elements to describe power price behavior.

The RisQuant models are statistical. However, they incorporate neither jump diffusion nor regime switching. The RisQuant models do incorporate mean reversion and persistence as secondary elements. Again, RisQuant models the power markets as a complex spark spread modulated by weather.