论文部分内容阅读
采用BP神经网络,可利用较少的输入参数建立地板辐射供暖系统热负荷预测模型,以大连市某超低能耗建筑为实测对象,根据实测的供暖期逐时热负荷数据建立了BP神经网络热负荷预测模型,并进行了改进。结果表明,采用基于多项式拟合改进的神经网络预测模型能够精确地预测一个单元未来24h的逐时热负荷,预测误差为5%左右。
BP neural network can be used to establish the thermal load forecasting model of floor radiant heating system with less input parameters. Taking a super-low-energy building in Dalian as the test object, BP neural network heat is established based on the measured hourly heat load data during heating period Load forecasting model, and improved. The results show that the neural network prediction model based on polynomial fitting can accurately predict the hourly heat load of a unit within 24h in the future with a prediction error of about 5%.