Risk Prediction Method for Renewable Energy Investments Abroad Based on Cloud-DBN

Abstract

There are many specific risks in renewable energy (RE) investment projects, and the incidences of these risk factors are fuzzy and uncertain. In different stages of a project’s life cycle, the main risk factors frequently change. Therefore, this paper constructed a cloud dynamic Bayesian network model (Cloud-DBN) for RE operation processes; it uses the DBN graph theory to show the generation mechanism and evolution process of RE outbound investment risks, to make the risk prediction structure clear. Based on the statistical data of observation nodes, the probability of risk occurrence is deduced to ensure the scientific nature of the reasoning process. The probability of risk being low, medium, or high is given, which is highly consistent with the uncertainty and randomness of risk. An improved formula for quantitative data normalization is proposed, and an improved calculation method for joint conditional probability based on weight and contribution probability is proposed, which reduces the workload of determining numerous joint conditional probabilities and improves the practicability of the BN network with multiple parent nodes. According to the 20-year historical statistical data of observation nodes, the GM(1,1) algorithm was used to extract the transfer characteristics of observation nodes, construct the DBN network, and deduce the annual risk probability of each risk node during the operation period of the RE project. The method was applied to the wind power project invested by China in Pakistan, and the effectiveness of the method was tested. The method in this paper provides a basis for investment decisions in the RE project planning period and provides targeted risk reduction measures for the project’s operation period.