| 摘要: |
| 为进一步提高FPSO工艺系统能耗预测准确率,避免设备供电不足或超压现象,研究以巴西海域某FPSO油气处理系统为研究对象,建立HYSYS工艺仿真模型；基于仿真数据构建了贝叶斯算法优化XGBoost(BO-XGBoost)能耗预测模型,获得了优化后的决策树数量、最大树深度、最小叶子节点权重和学习率；结合SHAP分析手段,定量探究了各影响因素对整体能耗模型的贡献值和重要性,得到特征在全局范围内对模型表现的总体影响。结果表明,相较于BP、RF和XGboost模型,改进模型的MAPE值平均降低11个百分点、7个百分点、5个百分点,优化后模型的分析框架对能耗拟合精度更高；系统中TEEDEA、Q、TEEREC和Pint是影响能耗的主要因素,现场应根据其工况特点和燃料气用气量合理分配阀门开度和压缩机入口压力。本研究的旨在通过精准识别FPSO油气处理系统能耗的关键因素,并运用贝叶斯优化算法优化的XGBoost模型(BO-XGBoost)实现对工艺系统能耗值的精确预测,有效解决FPSO设备供电不足或超压现象,为海上FPSO的安全平稳运行提供保障,同时提升能效水平,降低运营成本。 |
| 关键词: FPSO 贝叶斯优化 XGBoost SHAP 能耗预测 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| Research on Energy Consumption Prediction of an FPSO Oil and Gas Processing System in the Brazilian Sea Based on BO-XGBoost-SHAP |
|
Shen Dingjin1,2, Cui Yutao1,2, Wang Han1,2, Qiao Bailu1,2, Gao Guanqun1,2, Wang Lu1,2
|
|
1.China Offshore Engineering Equipment &2.Technology Co., Ltd.Shanghai
|
| Abstract: |
| To further enhance the accuracy of energy consumption prediction in FPSO process systems and avoid insufficient power supply or overpressure in equipment, this study focuses on an FPSO oil and gas processing system in Brazilian waters. A HYSYS process simulation model was established for this system. Based on the simulation data, a Bayesian Optimization-enhanced XGBoost (BO-XGBoost) energy consumption prediction model was developed, determining the optimized number of decision trees, maximum tree depth, minimum leaf node weight, and learning rate. Combining SHAP analysis, this study quantitatively explored the contribution and importance of various influencing factors to the overall energy consumption model, revealing the general impact of features on model performance across the board. The results indicate that compared to BP, RF, and standard XGBoost models, the improved model achieves an average reduction in MAPE by 11, 7, and 5 percentage points, respectively, demonstrating a higher fitting accuracy for energy consumption. Among the factors, TEEDEA, Q, TEEREC, and Pint are identified as the primary influencers of energy consumption. In practical operations, the valve opening and compressor inlet pressure should be reasonably adjusted according to operating conditions and fuel gas consumption. This study aims to precisely identify key factors affecting energy consumption in FPSO oil and gas processing systems and utilize the BO-XGBoost model to accurately predict energy consumption. By doing so, it effectively addresses issues of insufficient power supply or overpressure in FPSO equipment, ensuring safe and stable offshore FPSO operations while improving energy efficiency and reducing operational costs. |
| Key words: FPSO Bayesian Optimization XGBoost SHAP Energy Consumption Prediction |
