Zero-carbon building optimization design method based on deep reinforcement learning

Abstract

The invention belongs to the technical field of zero-carbon buildings, and particularly relates to a deep reinforcement learning-based zero-carbon building optimization design method, which comprises the following steps of: constructing four models: a parameterized model used for establishing information association between design parameters and an optimization target; based on a deep reinforcement learning-based design parameter and optimization target mapping model, rapid prediction of an optimization target can be realized; constructing a zero-carbon building optimization design model by using an NSGA-II algorithm, and obtaining a Pareto solution set; and the optimization design variable and optimization target collaborative design decision model is used for screening the Pareto solution set to make a further design decision. According to the mapping model constructed by the method, the calculation speed of a building optimization target can be increased, and the generalization ability of the building optimization target is improved; according to the decision-making method, on the premise that design variables are fully considered, all optimization targets can be evaluated and sorted, the decision-making range is reduced, the decision-making difficulty is lowered, and the design decision-making result is more comprehensive and scientific.

Description

technical field [0001] The invention belongs to the technical field of zero-carbon buildings, and in particular relates to a zero-carbon building optimization design method based on deep reinforcement learning. Background technique [0002] China has pledged to achieve carbon peaking around 2030 and achieve carbon neutrality by 2060. As one of the three major energy-using fields, the construction industry, realizing its zero-carbonization is the key to achieving the dual-carbon goal. The core of a zero-carbon building is to balance the building's own carbon emissions with the carbon consumption of renewable energy in the entire life cycle to achieve zero carbonization. Under the background of dual carbon goals, exploring zero-carbon building design methods will surely become the mainstream of future urban construction and development. [0003] The current architectural optimization design method is guided by the performance goal, and the parametric modeling and the optimiz...

AI Technical Summary

Problems solved by technology

Existing methods include icon-based decision-making methods such as Pareto scatter diagrams, multi-attribute decision-making methods for optimization goals, and self-organizing clustering neural network decision-making methods. The decision-making process of icon-based decision-making methods is too subjective, and the target value cannot be fully evaluated; multi-attribute decision-making The self-organizing clustering neural network decision-making method has a high degree of visualization, but it also fails to fully evaluate the optimization target values ​​in all Pareto solution sets.

Images