Among sub-items of energy consumption in public buildings, lighting sockets play an important role in energy-saving analysis. So, the energy consumption data quality of lighting sockets is important. However, limited by the initial cost of energy monitoring platform, it is difficult to install electricity meters covering all branches and to retrofit the incompact classification electricity branches, which results in a mixture of the lighting socket energy consumption and other components. In this study, a separation methodology is proposed. First, the abnormal data in the energy monitoring platform are cleaned and screened using a clustering algorithm. Second, the average outdoor air temperature partitioning model (OATPM) method and the k-nearest neighbor (KNN) clustering algorithm method are proposed for identifying and separating the abnormal data. These two methods have complementary advantages in the best applicable scenarios, including calculation accuracy and other aspects. The verification results for three buildings show that the relative error of this separation methodology is less than 15%. Finally, this paper presents the optimization parameters of the KNN method. Through this methodology, building managers need only historical data in an energy monitoring platform to separate the combined power consumption of the lighting sockets and air-conditioning online, independent of detailed information statistics.
The extraction and analysis of electricity consumption changing patterns are increasingly important, as they can guide in energy management and efficiency retrofitting. Consequently, it is necessary to extract the laws governing building electricity consumption characteristics. This method should be on an hour-scale and successfully applied online to various buildings. Under these conditions, the method should be as simple as possible to ensure excellent online applications. A matrix model method was developed based on the conventional K-nearest neighbour clustering algorithm, which introduced the concept of electricity intensity levels. This method used the slopes of the power consumption curve as the grading standard for the extraction and assessment of the electricity consumption laws. The validation results for seven different buildings with various functions and climate zones, including the mean absolute error, mean absolute percentage error, mean square error, root mean square error, and coefficient of variance, showed that this method met the aforementioned requirements. Moreover, this method can be used for power consumption prediction, which integrated a process of filtering historical data, leading to better accuracy and less data volume than that of other methods that use historical data for prediction. Practical application This paper proposed a matrix model method based on the conventional K-nearest-neighbour clustering algorithm, which introduced the concept of electricity intensity levels. This method was applied to various buildings online, which coupled the process of filtering historical data and flexible selectivity of models when used on different buildings. This method was used for assessing energy-saving potential, energy-saving retrofit priorities, and power consumption forecasting, which will benefit researchers and engineers.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.