The utilization of Empty fruit bunch (EFB) in energy production has increased in Malaysia over the last two decades. The EFB can be used as a solid fuel in a boiler system for heat and power generation. However, numerous safety and technical issues lead to a lower energy production rate. A holistic probabilistic risk analysis is developed using the Bayesian Belief Network (BBN) to reduce the risk in the boiler system. The Conditional Probability Table (CPT) indicates the influence strength between the parent node and child node in BBN. Due to scarcely available information on EFB boiler, elicitation from the expert’s opinion is vital. The formulation for boiler failures likelihood prediction that relies on experts’ perceptions was developed using the Weighted Sum Algorithm (WSA). A case study from BioPower Plant in Pahang was applied in this project. The model illustrates the relationship between the cause and the effect of the biomass boiler efficiency in a systematic way. Two types of analyses, prediction and diagnostic analysis, were performed. The results facilitated the decision-maker to predict and identify the influential underlying factors of the boiler efficiency, respectively. The result shows that the most important boiler failure factor is combustion stability. It agrees with experts’ experience that most biomass boiler failure was caused by EFB, which contains high moisture content that affects flame stability. The proposed formulation for expert opinions and perceptions conversion can be utilized for risk analysis to benefit the boiler and other infrastructure that relies on experts’ experience.