针对大型商场面向建筑整体冷负荷预测不能为商场各区域按需供冷提供合理控制策略的问题。通过研究商场不同区域冷负荷特点,采用灰色关联度分析法筛选影响商场不同区域冷负荷的关键影响因素,提出了基于注意力机制-长短期记忆(Attention-LSTM)神经网络分区冷负荷预测模型。LSTM网络充分考虑空调冷负荷与相关特征变量之间的非线性关系,特征注意力自主分析历史信息和输入变量之间的关系,提取重要特征,时序注意力选取LSTM网络关键时刻的历史信息,提升较长时间段预测效果的稳定性。以西安某大型商场建筑的冷负荷数据集为实验数据,实验结果表明所提模型相比于LSTM模型和CNN-LSTM模型误差指标MAPE和RMSE均有显著降低,R2明显增加且稳定0.99以上,具有较好的泛化能力和较强的稳定性。

Aiming at the problem that cooling load prediction of overall buildings for large-scale shopping malls cannot provide a reasonable control strategy on demand for various areas of the shopping mall. According to the characteristics of the cooling load in different areas in the mall, the grey correlation analysis method was used to screen the key influencing factors of cooling load in different areas of shopping malls, and a district cooling load forecasting model based on attention long-short term memory (Attention-LSTM) neural network was proposed. LSTM network fully considers the nonlinear relationship between air conditioning cooling load and related characteristic variables. Characteristic attention autonomously analyzes the relationship between historical information and input variables to extract important features. Time-series attention selects historical information at key moments in the LSTM network to improve the stability of the forecast effect

TP391.9

国家重点研发计划(新型建筑智能化系统平台技术)项目(2017YFC0704100)