基于数据驱动的建筑电力灵活性潜力评估方法

Data Driven Evaluation Method of Building Power Flexibility Potential

摘要: 至2021年，建筑物占全球能源消耗量的40%，在匹配发电和需求响应方面可以发挥重要作用。为应对能源供给侧的要求的当下，短期建筑电力灵活性潜力评估方面仍缺乏研究。随着以新能源为主体的新型电力系统发展，规模化灵活需求侧互动响应资源挖掘成为必然。针对建筑缺少需求响应历史数据时的灵活性潜力分析问题，本文提出了基于多源无监督域适应的建筑电力灵活性潜力评估方法，充分利用已参与响应项目建筑的历史数据来评估建筑电力灵活性潜力。首先，基于仿真模拟方法来构建多个典型办公建筑模型并量化建筑的电力灵活性潜力；其次，构建基于CNN-GRU的特征提取器用于提取多源时间序列数据的时间特征；在此基础上，通过特征提取器和具有softmax输出的域分类器的对抗域自适应来找到源域和目标域之间的域不变特征，更有效地利用源域标签数据，最终实现了缺少建筑历史响应数据时的灵活性潜力评估神经网络的训练；最后，通过算例分析，验证了本文所提方法的有效性。

Abstract: By 2021, buildings account for 40% of global energy consumption and can play an important role in matching power generation and demand response. Buildings should meet the requirements of en-ergy supply side, and there is still a lack of research on short-term energy flexibility prediction. With the development of new power systems with new energy as the main body, large-scale flexible demand-side interactive response resource mining becomes inevitable. In view of the problem of building flexibility potential analysis when there is no demand response historical data, this paper proposes a building power flexibility potential evaluation method based on multi-source unsuper-vised domain adaptation, which makes full use of the historical data of the buildings that have par-ticipated in the response project to evaluate the building power flexibility potential. Firstly, a gen-eral method based on simulation is used to quantify the energy flexibility of zero energy consump-tion buildings and determine the potential of typical types of ZEB; secondly, a multi-source domain adaptation model is designed for the time series. The feature extractor based on cnn-gru is used to extract the time features of the input time series data, and the domain invariant features between the source domain and the target domain are found through the anti domain adaptation of the fea-ture extractor and the domain classifier with softmax output; finally, the effectiveness of the pro-posed method is verified by an example