数据驱动的固定拓扑结构换热网络优化改造方法

获得经济可行的换热网络改造方案目前仍然是一个巨大的难题。传统的改造方法严重依赖拓扑修改，这往往会增加改造时间和投资成本。因此，本文将围绕固定拓扑结构的换热网络改造展开研究，提出了一种基于数据驱动的高效改造方法。该方法通过建立性能模拟模型，模拟计算换热器在不同性能参数时的公用消耗与温度分布，以此获取大量的驱动换热网络改造的数据。然后通过BP神经网络预测模型与遗传算法，求解得到以最大节能量为目标的改造方案。案例研究表明，改造后的换热网络可以通过较小的投资获得较大的节能效益，与文献相比，单位费用节能量提高了51.4%；与同为固定拓扑结构的灵敏度分析相比，可以快速地获得经验规则法无法获得的改造方案，验证了改造方法的经济实用性与高效性。

Data-driven optimization retrofit method with fixed topology structure for heat exchanger network

Finding cost effective retrofit for heat exchanger networks remains a challenge. Traditional retrofit methods rely heavily on topological modifications, which often increase retrofit time and investment costs. Therefore, this paper focuses on the heat exchanger network retrofit with a fixed network structure, and a data-driven efficient optimization method was proposed. A performance simulation model was established to simulate the utility consumption and temperature distribution of a heat exchanger under different performance parameters, so as to obtain a large number of data for the optimization of the heat exchanger network. Then back propagation neural network prediction model and genetic algorithm were used to solve the retrofit problem with the goal of maximum energy savings. Case studies showed that the retrofitted heat exchanger network can obtain greater energy-saving benefits through smaller investment. Compared with the literature, the energy savings per unit cost increased by 51.4%. A modification scheme that cannot be obtained by the empirical rule method can be quickly obtained compared with the sensitivity analysis of the same structure. Moreover, the effectiveness and economic practicability of the proposed method were demonstrated.