基于遗传算法的建筑热阻容模型辨识及室内温度预测
Building thermal resistance-capacitance model identification andindoor temperature prediction based on genetic algorithm
摘要：
为了准确预测建筑物室内温度,采用集总参数法构建了建筑热阻容(RC)模型,并推导出了相应的微分方程组,使用遗传算法对模型进行了辨识。RC模型测试集温度平均绝对误差为0.14 ℃,R2为0.99。比较了RC模型与2种黑箱模型对室内温度的预测结果,RC模型预测结果更准确。讨论了灰箱模型和黑箱模型的适用条件,建议根据需求和数据情况并考虑模型的复杂度、预测精度、解释性等因素,选择合适的建模方法。
Abstract：
In order to accurately predict the indoor temperature of buildings, this study adopts the lumped parameter method to construct a building thermal resistance-capacitance (RC) model. The corresponding differential equations are derived, and the model is identified using the genetic algorithm. The RC model’s test set indicates an average absolute error of temperature of 0.14 ℃ and a R2of 0.99. The prediction results of indoor temperature between the RC model and two black box models are compared, and the prediction results of the RC model are more accurate. This paper discusses the applicability conditions of gray box and black box models. It is recommended to select an appropriate modeling method based on the needs and data conditions, and considering the complexity, predictive accuracy, interpretability and other factors of the model.
Keywords:indoor temperature prediction; genetic algorithm; lumped parameter; thermal resistance-capacitance model; gray box model; black box model