The global low-carbon transition is driving refining enterprises to shift from oil-dominant production toward diversified production, requiring the collaborative allocation of refined oil and new energy. However, logistics systems face significant challenges arising from nonlinear demand fluctuations and external disturbances, which traditional static approaches fail to address effectively. To this end, this study proposes a multi-party collaborative framework for multi-energy logistics systems. Specifically, a hybrid demand forecasting model integrating variational mode decomposition and bidirectional long short-term memory is developed to capture demand fluctuation characteristics. Based on the forecasting results, a multi-product, multi-modal rolling robust simulation model is constructed to dynamically balance economic benefits and system resilience under uncertainty. The proposed framework is validated through a case study of a large Chinese petrochemical enterprise, and a multi-dimensional evaluation method incorporating resilience, economic, environmental and energy indicators is proposed. The results show that: (i) compared with the deterministic scheme, the proposed simulation scheme increases operating costs by 3.64% under high volatility scenarios, while improving service level by 3.18%, reducing average stockout duration by 29.5%, and decreasing stockout quantity by 46.15%; (ii) the optimized energy product structure achieved by the model reduces carbon emissions by 7.7% and increases calorific value by 7.6%; and (iii) the study reveals integrated refinery and new energy configurations enabled by multi-party collaboration, providing quantitative support for enhancing system resilience. Based on these findings, policy implications are proposed to promote the development of energy logistics systems.