The LOPA is a semi-quantitative assessment method. If the assessment of risk is not precise, then a reasonable assessment may not be made. Human factors affect assessment results because expert judgments are involved in the LOPA assessment. This paper proposes an improved layer of protection analysis (LOPA) called the cloud model layer of protection analysis (CM-LOPA). A cloud model (CM) is used to improve the LOPA. Expert judged risk is processed by the CM, the quantified risk is presented as the cloud model risk (CMR), and the algorithm of the CMR is developed. A case study of gas leakage in a biomass gasification station is developed based on the CM-LOPA. A risk-matrix-based LOPA is introduced and compared with the proposed method. Assessment and comparison results show that the randomness and fuzziness of the CM make assessment results obtained by CM-LOPA more precise and reasonable. (C) 2017 Elsevier Ltd. All rights reserved.