Solar Developers' Losses due to Sunshine Duration Shortfall and Marketability of Sunshine Insurance: A Loss Distribution Approach

Abstract

Photovoiltaic energy is a rapidly increasing renewable energy source in Korea. As of 2014, it constituted 4.7% of the renewable energy mix, an increase of 58.9% from the previous year (KEA, 2014). To maintain this trend, it is essential to minimize solar generation volatility and other risks related to solar energy development. Solar developers face market risks (SMP and REC price, FIT and PPA availability, etc.), technical risk (system performance, tests reliability, infrastructure level, etc.), weather/climate risk (extreme weather events, natural disasters, solar radiation/sunshine shortfall, etc.). Among those, this paper will focus on sunshine duration shortfall, since it is one of the factors that systematically affects solar energy developers and has high correlation with the amount of solar energy produced (as much as 0.8 according to some empirical studies in South Korea). This paper will analyze the typical risk mitigation tool, insurance against potential losses, which can be categorized as business interruption insurance. Despite of the high correlation between generated energy output and sunshine duration hours, only a few Korean insurance companies offer solar insurance, and with little success so far. To examine the potential demand, this paper will first compute the number of cities that have faced sunshine shortfall on yearly basis from 1990 to 2015. This paper created two plausible criteria defining sunshine shortfall: 1) when an average annual sunshine duration hour in a particular city and year falls more than 10% of the historical average (scenario 1) or when the same variable drops over 20% of the historical average (scenario 2). The former scenario will be discussed in detail, whereas the latter scenario will be used for suggestion part of this paper. Based on the frequency of sunshine shortfall events and solar developers revenue estimation, this paper will calculate the severity of the losses due to sunshine shortfall. Based on this information, the paper will derive the most suitable probability distribution describing the frequency and severity of the events, respectively. Then, it will check for their statistical significance by using Kolmogorov-Smirnov and Anderson-Darling tests. After selecting the distributions that match the data, this paper will estimate the parameters for each case using Maximum Likelihood Estimation and Bayesian Estimation methods. The two methods will be applied separately on the same data set in order to improve the credibility of the results and provide richer information on the topic. Then, for each combination set of the estimated parameters, this paper will apply Loss Distribution Approach (LDA), the method often used in the evaluation of operational risk, by summing yearly risk and conjugating the frequency and severity distributions to obtain 50th, 75th, 90th, 99th and 99.9th percentile of the aggregate loss distribution. Finally, this paper will perform a simulation on the LDA results 30 times for each paired outcome of frequency and severity to use the average value of each set for risk interpretation. This paper intends to assess the attractiveness of sunshine shortfall insurance both from the perspective of solar developers and insurance companies. By calculating different percentiles of Value at Risk with the use of Loss Distribution approach, this paper will try to answer why sunshine (or solar radiation) shortfall insurance is still an infantile industry and why the current providers have to offer it at high premiums. This paper is the first academic research in Korea that answers 1) how much volatility sunshine duration has

2) based on the synthesis of current sunshine shortfall contracts, what the following revenue loss due to sunshine shortfall is

3) how the result change if we considered more realistic (by incorporating Renewable Energy Certificates sales revenues inside) calculations of sunshine shortfall

4) whether climate change affects average yearly sunshine duration hours and how it can be translated into a higher risk for insurance companies

5) whether it is possible to make sunshine duration insurance affordable to more solar developers. There are four important findings in this paper. First, risk value exceeds insured value at high percentiles. Consequently, it is reasonable to assess the realizable risk at 50th and 75th percentiles of the aggregate probability distribution. Second, when rescaled at per KW level, fixing the base year at 1990th level creates risk value 2.2 to 2.6 times higher than with moving base year (based on the sunshine duration average of directly preceding 10 years). Third, by incorporating RECs in solar developers risk valuation, this paper concludes that the risk will soar by 1.9 to 2.3 times. Finally, although this paper focused on scenario 1, which defined sunshine duration shortfall as any shortfall below 10% of the historic average, it also suggested an alternative, where sunshine shortfall is defined as any shortfall below 20% of historic average. The results showed that an insurance product following the latter insurance structure could reduce the premium much more than in the former insurance structure. In the less riskier insurance structure case, the ratio of fixed costs to insurance premiums and the share of insurance premiums in energy sales revenue were close to the NREL(2010) indicators, making it more realizable.