Neural networks have been applied in analyzing the influencing factors of electricity consumption in residential households nationwide and have achieved good results in electricity consumption prediction. However, the invisibility of the decision-making process of the neural network system leads to its inability to get a good solution in explaining the influencing factors. It was found that the interpretability of BP neural network was 0.902, compared with 0.847 of FEM model, and the SHAP model gave the marginal contribution of each feature. Also, among the features affecting electricity consumption, household water consumption, retail sales of consumer goods, and education expenditure have important values for residential household electricity consumption. In this paper, we discuss the application of neural networks in the factors affecting residential household electricity consumption from the perspective of interpretability, and innovatively introduce the SHAP interpretation method to study the important indicators in the rating, which is important for the research in the field of residential household electricity consumption.