重质燃料油在线优化调和系统建设所面临的挑战

Challenges of On-line Optimizing Blending System Building of Heavy Fuel Oil

分析重质燃料油在线优化调和系统的建设内容及所面临的挑战。基于现有指标预测模型不适用于重质燃料油调和的现状,建议基于支持向量机等工具建立具有自适应性的非线性调和指标预测模型,有效利用新数据不断提高模型预测精度。以基于支持向量机的调和指标预测模型为限制条件,建立重质燃料油优化调和模型,并探讨模型求解算法。以自启发式算法所搜索到的批量次优方案为初值,通过传统非线性规划算法或线性规划+递归算法对其进一步寻优,可提高获得全局最优解的概率。设计具有自适应性的优化调和流程,通过人机结合的方式,在"疑似"局部最优解的邻域内快速提高预测模型精度,促进在线优化调和系统尽快起效。

The construction contents and facing challenges of on-line blending system of heavy fuel oil were analyzed.In order to sustainably improve the prediction accuracy with new data,the prediction model of non-linear blending index with adaptability based on mathematical tools,such as support vector machine(SVM),was suggested.An optimization blending model of heavy fuel oil with constraints based on SVM was built,and the corresponding solving algorithm was discussed.Taking a batch of inferior optimized solutions searched by heuristic algorithm as initial values,and by further optimization with traditional non-linear programming method or linear programming plus recursive algorithm,the global optimization solution could be obtained.An optimization blending procedure with adaptability was designed.By this human-computer cooperation method,the prediction accuracy would be improved quickly,thus the online blending system would take effect rapidly.