Bayesian updating of solar resource data for risk mitigation in project finance

摘要

Project finance is based on the future cash flow of projects. Ensuring that the expected revenue of projects will cover the debt and equity obligations issued by lenders and shareholders is crucial. The uncertainty of solar resources is among the highest, and it causes fluctuations in the future cash flow of solar photovoltaic (PV) projects. To reduce this uncertainty, several methods such as measure-correlate-predict (MCP) analysis, have been applied. However, MCP is an oversimplified linear regression method that disregards the difference between the parameters and conditions of different hours throughout a day; hence, it cannot provide accurate and reliable results. Here, we propose a methodology based on Bayesian updating, which is a robust probabilistic approach to reduce the aforementioned uncertainty. We use the Metropolis-Hastings algorithm and four years of onsite measurements to obtain the posterior distribution of hourly solar resource data. Then, we demonstrate that our proposed method improves the reliability of indices of project finance deals by applying it to a 10 MW solar PV project. To facilitate decision-making in determining the leverage for a project finance deal, particularly in the case of material default, we introduce conditional value-at-risk (CVaR) for the distribution of the debt service coverage ratio (DSCR). We calculate DSCR in three cases: applying MCP and the Bayesian updating method for risk mitigation and without using any risk reduction approach. The results demonstrate that higher financial leverages can be selected by choosing a rational threshold amount for CVaR that corresponds to the boundary of material default.