Non-Intrusive Load Monitoring Based on Data Correlation for Distribution Network

Non-intrusive load monitoring (NILM) has significant supporting effect on demand response. However, the mini-watt equipment, continuous operation equipment and variable state equipment have brought great difficulties to the identification accuracy. This paper proposes a method to identify load components based on the data correlation. This method is constructed via three steps: first, a physical electrical constraints model is applied to abstract the relationship between load customers and distribution network; second, a stepwise method with cross validation is used to select the most relevant features; and third, the load components are explained by the characteristics model and solved by combinatorial optimization method. Finally, the accuracy of the proposed algorithm is verified by comparison against the NILMTK algorithms through detailed simulations over high frequency engineering data.