When some household appliances are running, the electrical quantities such as active power and reactive power will change slowly for a long time, which will affect the feature extraction of event-based non-intrusive load identification, resulting in low accuracy of load identification results. In view of this situation, a feature extraction method is proposed. Its core idea is that there is a certain law in the change of active power and reactive power in a very short time, and the curve can be fitted to speculate the values of these electrical quantities in the next time, so as to eliminate the influence. Based on this, a feature extraction method applied to non-intrusive load identification is realized. The test results show that this method can effectively improve the accuracy of non-intrusive load identification feature extraction data, and has good application value.