中文摘要：

针对大规模变工况流程模拟中初值要求高、收敛性差的问题，提出了一种结合同伦思想与回溯搜索方法的回溯同伦法（HBM）。该方法利用工况变量作为同伦参数，提高了同伦辅助问题的可解性。采用回溯法自动搜索同伦参数并作为非线性方程组求解器的外壳，降低了求解器对初值的要求。在Aspen Plus中将回溯同伦法与内部求解算法结合对乙烯分离过程变工况算例进行了模拟，结果表明HBM能够扩大收敛范围，可有效达到问题的物理边界。同时，使用回溯法搜索工况参数能保证找到同伦工况点，帮助分析过程瓶颈。

英文摘要：

A homotopy-based backtracking method （HBM） for large-scale process simulation with load variations was proposed to improve the convergency of nonlinear algebraic equation solver. The fluctuated load variable was regarded as a homotopy parameter to make the auxiliary homotopy problem easily converge. In this paper, the homotopy parameter is searched by a backtracking method, which greatly improves the searching efficiency. With the Open Object Model Framework （OOMF） script language, the proposed HBM can be connected to solvers in Aspen Plus for the process simulation. The eraluation of the HBM were conducted via the simulation calculation of a large scale ethylene separation process with load variations. It is shown that the proposed HBM greatly enhances the convergency performance. It consistently converges to an optimal solution until it encounters an unsurpassable operation boundary. The backtracking method is also effective in searching homotopy parameter, solving auxiliary homotopy problem and analyzing flowsheeting bottleneck.