Related to this, I discussed a paper by Richard Green and Iain Staffell “How Large Should a Portfolio of Wind Farms Be?”at the Conference on The Economics of Energy Markets organized by Toulouse School of Economics between Jan 17-18, 2013. The paper models the performance of –not-yet-established– wind farms in the UK for 2012-2020. The main focus is to exhibit the potential benefits of spreading windmills over a large area. They make use of a full market model and conjecture the revenues and the costs of 32,767 postfolios of wind farms. They argue that from the point of making an investment, measuring expected annual profits and their variance can give the most relavant information for investors.
Following this, the authors calculate the efficiencies of these portfolios based on the distance between an efficiency frontier (a frontier built by the portfolios where it is impossible to raise the expected profits and at the same time reduce the variance) and an optimal portfolio, which has the highest profit and lowest variance (see the figure). The best of these portfolios can be observed to the south-east of the figure, which are the ones that give the highest mean annual profits and the lowest deviations.
Their results show that optimal portfolios are contained in no more than 4 regions (see the table), suggesting that the benefits of having a portfolio of windfarms is an attainable target until 2020.
One assumption in the analysis is that the cheapest available stations are assumed to be always available, meaning that they will be capable of meeting any demand. Hence in case you are an investor, you should not be worried about things like first movers since these sites will always be capable of supplying the land and others the investment would require. I find this assumption rather debatable, especially when a longer period of time is considered.
Suppose the most profitable portfolios are taken and hence no more available. This will naturally result in a north-west shift in the efficiency frontier. What will happen for sure after the shift is that the investors will be left with optimal portfolios that would contain a higher number of sites, possibly making it quite difficult to attain the most profitable portfolios. This may imply that the time period after 2020 could be quite rough for investors. Hence, it will be quite interesting to see some dynamics in the analysis.
As building a full market model to assess the portfolios is a bit of a challenge, the authors ask also whether it would be sufficient to assess only the expected annual output and its variance. To find an answer, they look at the correlation between the two measures (efficiency scores) and find that it is only 0.103 (see the figure). This is to say that the most profitable portfolios will not likely to be the ones that would be most productive. This may make the crowd go wild and favor the measure based on the profits, but maybe we should not be so fast in arriving this conclusion. Since the optimal portfolio analysis coming out of the full market model is based on many details, this strongly necessities a careful sensitivity analysis (The sensitivity of the results is mentioned by the authors as well. However, why there is no sensitivity analysis is a mystery).
All in all, portfolio efficiency measured by profits make things easier from an investor’s point of view. However, to be confident about this, it is crucial to know and show that the results are not sensitive to some crucial details, which most possible have the tendency to variate in time. Lastly, incorporating dynamics, areal limitations and some strategic behavior into the analysis will take it a step further and make things even more interesting, though not less challenging.
You can view this post also @Reconhub