Managing a large number of participants in the Delphi survey encounters many challenges, such as heterogeneous data, high cost, dynamic quality, and various uncertainties, etc. This paper proposes a novel large-scale group Delphi method with heterogeneous decision information and dynamic weights to address these challenges. Heterogeneous data with various uncertainties, including exact numbers, interval numbers, normal cloud models, linguistic terms, and linguistic expressions, are converted into normal cloud models. Thus, both magnitudes and uncertainties of data can be preserved and propagated throughout the whole calculation process. Furthermore, this study also presents a data preprocessing technique consisting of four steps to discard invalid data and low-quality inputs. Moreover, a novel dynamic weights assignment algorithm is proposed to determine the weights for participants. Afterwards, detailed procedures and formal algorithm are developed. A practical application is presented to demonstrate the effectiveness, applicability, and advantages of the proposed methodology with examples and detailed comparative analysis. The application shows that the decision result becomes more stable and consentaneous. The comparative analysis indicates that the proposed methodology overcomes several challenges. Therefore, the proposed method can provide novel ideas for large-scale group decision making with heterogeneous data and various uncertainties in a dynamically changing decision environment. Eventually, this study contributes to several fields, such as the Delphi technique, group decision making, modeling of uncertainty, and heterogeneous data representation and processing, etc.