For the theoretical research and evaluation of nonintrusive load monitoring (NILM) technology, a large-scale and diverse labeled dataset that includes multiple user types and conforms to the normal usage habits of users is required as a basis. While on-site (partial) intrusive measurements can be used to obtain (partial) labeled data from real world users over a period of time, it is costly to establish a fully labeled dataset this way that meets the number and diversity of users required. In this paper, taking into account the impact of the factors from geographical, seasonal, user type and others on the load composition and usage habits, as well as the different power variation patterns due to the different working mechanisms of appliances, a regional multi-user load power dataset simulation method that relies on the appliance usage behaviors and power consumption traces of the real-world households and appliances is proposed. The preliminary test shows that the fully labeled dataset generated by the suggested method is reasonable for NILM research and evaluation.