Non-Invasive Online Abnormal Load Identification Method Based on AOA-SVM

In recent years, the requirements of power system for load monitoring in the substation area are increasing. Aiming at the problem of abnormal load identification under the background of unknown household appliances load, this paper proposes a non-invasive on-line identification method for charging load of electric vehicles based on AOA-SVM. Firstly, the multi-dimensional features are sorted and screened by reliefF algorithm to get the optimal feature subset. Then, the arithmetic optimization algorithm (AOA) is used to optimize the parameters c and gamma of the support vector machine, and the AOA-SVM classifier is obtained. Finally, the anti-jitter mechanism is used to filter out the false alarm events and accurately identify and detect the charging load of electric bicycles at home. The real data obtained in the low-voltage area are used for experimental verification, and the experimental results prove that the method in this paper has good applicability and accuracy under the background of unknown household appliance load.