摘要
提出求解数据校正问题的两层次变换算法。在对已测量数据和未测量数据是否可被校正和估计进行理论分析的基础上，采用投影矩阵进行数据分类和协调计算，提出采用投影矩阵对约束方程进行两个层次的变换算法。该算法针对目前通用的由Ｃｒｏｗｅ提出的算法在求解未测数据的估计值时很可能遇到对奇异矩阵求逆从而无法进行正确计算的缺点，在求解校正值和估计值时对约束方程进行深浅不同的两个层次的变换，可缩小解题规模和杜绝出现对奇异矩阵求逆的情形，提高了计算效率和计算的可靠性。
A twolevel transformation algorithm for data reconciliation is presented. Based on the theoretical analysis of measured quantities' redundancy and unmeasured quantities' observability, the concept of matrix projection is used in the work of data classification and reconciliation. A method of transforming the constraint equations on two levels by matrix projection is proposed. If the universal algorithm presented by Crowe is used, the occurrence of a singular matrix in matrix inversion may cause calculation error in the solution procedure of estimated values. To overcome this obstacle, the constraint equations are transformed on two different levels in the solution procedure of reconciled values and estimated values. The algorithm is practical for online data reconciliation, because it can reduce the dimension of problems to the largest extent, avoid the matrix inversion error and improve the computational efficiency and reliability.
出处
《石油炼制与化工》
CAS
CSCD
北大核心
1999年第6期56-60,共5页
Petroleum Processing and Petrochemicals
