Aiming at the complex and changeable construction risk assessment of photovoltaic power generation projects, the Leaky Noisy or Gate model is used to integrate historical experience and expert knowledge to determine the Bayesian network parameters .The Bayesian network model of photovoltaic construction risk is established. At the same time, the Netica simulation analysis is used to effectively predict and diagnose the construction risks and identify the key risks and sensitive risks, so that the managers can make accurate decisions and control the risks. In order to verify the effectiveness of this method, the Z project is taken as an example for modeling and simulation; The results show that the construction risk level is medium, which is consistent with the actual situation.