A method and device for predicting thermal performance and extracting mechanisms of a water-based thermally activated building system

Abstract

The application provides a water-based thermal activated building system thermal performance prediction and mechanism extraction method and device, and the method steps comprise: constructing a system benchmark model; defining characteristic variables of the water-based thermal activated building system, and determining threshold values of the characteristic variables; performing sampling on the characteristic variables to generate a characteristic variable sample library; building and configuring a random finite element calculation program to perform parameterized solving on samples in the characteristic variable sample library; outputting a first result based on thermal performance index data post-processing to obtain a second processing result based on the thermal performance evaluation index; establishing a data set containing the characteristic variable sample and the thermal performance evaluation index; obtaining an optimal performance prediction base model through a machine learning model combined with a Bayesian hyperparameter optimization method; and applying an interpretable artificial intelligence method to perform explanatory analysis on the optimal thermal performance prediction model to extract the action mechanism between the characteristic variables and the thermal performance. Compared with the prior art, the application has the advantages of simplicity, efficiency and strong generalizability.

Images