Load components reflect the amount of contributions that different building functional elements have made to the generation of total load during building operation. It will be helpful for a more accurate load prediction, and also for indirect evaluation on the thermal performance of building envelopes if the load components can be fully made out. However, in practical engineering, the data of load components cannot be obtained by field test. In this paper, benefiting from the different morphological features of load component profiles, a method based on morphological components analysis (MCA) is proposed to disaggregate cooling load into four parts with specific physical significance: temperature difference load, TCL; solar radiation load, SRCL; fresh air load, FACL; and internal load, ICL. Firstly, non-negative dictionary learning algorithm is introduced to learn the theoretical morphological feature of each load component adaptively. Secondly, non-negative sparse coding algorithm is adopted to disaggregate the time series of total load, and finally the load components are extracted from total load. Two cases with different settings using simulated data have been implemented. The results demonstrate that the proposed method can realize the load disaggregation effectively and accurately, where the accuracy index, Accuracy, is above 85% under different day attributes.
Read full abstract