A method for identifying specific load state of distribution network

Non-intrusive load monitoring (NILM) is a key component in the identification of smart electricity usage. Since there are many types of loads connected to the distribution network at the same time, that of them have variable frequency functions and do not have the characteristics of constant power, the existing load identification methods that focus on households are difficult to directly apply to the distribution network. Aiming at the load characteristics of the distribution network, this paper selects the elevator as a typical load to carry out a load identification experiment, and uses the measurement data conforming to IEC 61000-4-30 as input. The goal is to identify whether the elevator is in operation. In order to eliminate the computational pressure caused by irrelevant features, a differential feature extraction method based on Pearson's correlation coefficient is proposed, combined with a convolutional neural network, to realize elevator load status identification in an environment with multiple unknown loads. The results of using the measured data show that the method only needs a small number of samples to identify the elevator running state with complicated changes in operating power consumption, and the calculation accuracy is higher than that of the traditional machine learning method.