Abstract: The fine-grained energy consumption monitoring is one of the important links of smart grid construction. As an energy consumption monitoring method, non-intrusive load monitoring can deeply analyze the fine-grained load components of users, which is of great significance to the power optimization of users, and also has the characteristics of fast implementation and low cost. Firstly, the hidden Markov model is improved by relaxing the assumptions of the model, and then the household power load is modeled based on the improved hidden Markov model. Finally, the improved Viterbi algorithm is used to solve the optimal state transition sequence of the load equipment, and then the power consumed by each equipment is calculated. The experimental results show that the proposed improved algorithm not only has high accuracy, but also has good stability. At the same time, the decomposed power curve is more consistent with the actual power curve, and has good effect.