转炉煤气柜位的多输出最小二乘支持向量机预测

Multi-output least squares support-vector-machine for level prediction in Linz Donaniz gas holder

针对钢铁企业转炉煤气系统中煤气柜位的预测问题,提出一种一致T型灰色关联分析确定柜位的主要影响用户,避免了冗余输入因素,降低了预测模型的复杂度.根据多个不同煤气柜同时并网运行的情况,提出多输出最小二乘支持向量机回归算法来建立柜位预测模型.该算法采用等式约束,通过最小化所有输出的单一和整体拟合误差,将其转换为求解一系列线性方程组,得到模型回归函数的权系数和偏置公式表示.现场数据仿真实验结果表明所建预测模型的有效性和实用性,为制定煤气调配方案提供了合理指导.

To predict the level of linz donaniz gas（LDG） holder in iron and steel enterprises, we propose a consistent T＇s grey relation analysis to determine influential users, thus avoiding redundant inputs and reducing the model complexity. Considering the parallel connection of gas holders in operation, we employ the multi-output least squares support-vector- machine regression algorithm with equality constraints to develop the prediction model. The weighting factors and the bias of the model are determined by solving a series of systems of linear equations obtained from the minimization of the individual and overall output fitting errors. Simulation results using field data show the effectiveness and applicability of the proposed model, providing a reasonable guidance for the LDG resource allocation.