A framework for baseline load estimation in demand response: Data mining approach

In this paper we propose a framework of customer baseline load (CBL) estimation for demand response in Smart Grid. The introduction of demand response requires quantifying the amount of demand reduction. This process is called the measurement and verification. The proposed framework of CBL estimation is based on the unsupervised learning technique of data mining. Specifically we leverage both the self organizing map (SOM) and K-means clustering for accurate estimation. This two-level approach efficiently reduces the high dimension of the input vectors into two dimensional output using SOM, and then this output vectors can be efficiently clustered together by K-means clustering. Hence we can easily find the load pattern that is expected to be similar to the potential load pattern of the day of demand response (DR) event. To validate our method we perform large scale experiments where the building complex power consumption is monitored by 2,500 smart meters. Our experiments show that the proposed technique outperforms a series of the day matching methods. Specifically, we find that the root mean square error is reduced by 15-22% in average, and the mean absolute percentage error is reduced by 15-20% in average as well.