大型建筑改造过程中多尺度多目标智能优化关键技术研究

Building retrofit can improve building performance with respect to energy efficiency, comfort level, and safety. Intelligent design of optimal retrofitting strategy can help enhance buildings’ automation level and reduce cost regarding to design by hand, so that retrofitting projects, including a number of buildings, can be designed with a global optimization approach. As buildings belong to large-scale complex systems, optimal decision making can be regarded as a dynamic multi-objective optimization problem, which includes many challenges regarding to modeling and optimization. Therefore, in this project we plan to study modeling and optimization methods for this application. We mainly focus on how to predict the dynamic model (parameters and objective), how to ensure diverse convergence of multi-objective optimization algorithm, and how to design a dynamic multi-objective optimization algorithm. Accordingly, we firstly study data-driven modeling in an inverse approach. Secondly, we study a multi-objective optimization method based on neighborhood field, which can improve convergence speed and diverse distribution. Thirdly, we study a dynamic multi-objective optimization based on a receding horizon method, and study a multi-variable optimization method for optimal retrofit planning. As a result, intelligent design of retrofitting plan could be expected. A framework could be expected for dynamic multi-constraint and multi objective models. Convergence and diversity is expected for the improved multi-objective algorithm. Intelligent decision making could be expected for optimal cost, energy saving, and earning.

建筑物改造可有效提高能效、舒适和安全等各项性能。建筑改造智能优化设计，有利于提高建筑业自动化水平，降低人力设计成本，实现大型建筑改造项目的统筹优化调度。因建筑物是复杂大系统，建筑物改造的决策过程是多约束的动态多目标优化问题，面临建模和优化两方面的挑战。因此，本课题拟围绕这两方面开展研究。针对三点关键科学问题：如何预测动态模型（参数或目标值）？如何保证多目标优化算法的收敛性和多样性？如何设计动态多目标优化算法？分别研究：1）基于数据驱动的逆向建模方法，及多尺度多目标模型的动态特性；2）基于近邻场的多目标优化算法，及其收敛性和多样性；3）基于滚动优化机制的动态多目标优化算法，及其面向建筑改造决策问题的多变量优化方法。解决建筑改造智能设计中建模和优化问题，构建一个完备的多约束多目标动态模型的框架，提出了兼顾收敛性和多样性的多目标优化算法，实现改造成本、节能量和经济收益等多目标优化的智能化决策。

建筑物改造可有效提高能效、舒适和安全等各项性能。建筑改造智能优化设计，有利于提高建筑业自动化水平，降低人力设计成本，实现大型建筑改造项目的统筹优化调度。因建筑物是复杂大系统，建筑物改造的决策过程是多约束的动态多目标优化问题，面临建模和优化两方 面的挑战。因此，本课题拟围绕这两方面开展研究。针对三点关键科学问题：如何预测动态模 型（参数或目标值）？如何保证多目标优化算法的收敛性和多样性？如何设计动态多目标优化 算法？分别研究：1）基于数据驱动的逆向建模方法，及多尺度多目标模型的动态特性；2）基于近邻场的多目标优化算法，及其收敛性和多样性；3）基于滚动优化机制的动态多目标优化算法，及其面向建筑改造决策问题的多变量优化方法。分别取得以下成果：1）时空数据驱动的链式ESN算法，及多时间尺度ESN算法；2）近邻场的网络特性分析，及基于代理模型的近邻场算法；3）众包机制下动态博弈理论及多目标优化算法。解决建筑改造智能设计中建模和优化问题，构建一个完备的多约束多目标动态模型的框架，提出兼顾收敛性和多样性的多目标优化算法，实现改造成本、节能量和经济收益等多目标优化的智能化决策。