Non-intrusive Extraction and Forecasting of Residential Electric Vehicle Charging Load

Abstract

As with increasing growth of electric vehicles in the demand side, how to accurate identify the electric vehicle charging behavior is crucial to achieve stable operation of the power grid and strengthen possible ancillary services. Due to the limited direct monitoring data, it is of great significance to extract electric vehicle charging load with the aggregated smart meter data from the household perspective. In order to deal with that, this paper presents a novel framework to realize the non-intrusive extraction and forecasting of residential electric vehicle charging load. Firstly, the Factorial Hidden Markov Model (FHMM) algorithm is used to extract the charging load of the residential electric vehicles. The appropriate number of hidden states is selected by the iterative k-means method to establish the FHMM model of simultaneous operation of multiple devices. Then, the Long-Short Term Memory (LSTM) deep learning algorithm is used to forecast the electric vehicle charging load of residential electric vehicle in a short time period. Real smart meter data is used to test the proposed method and the result shows its feasibility for significant performance in the extraction, as well as the forecasting.