This study combines the electricity consumption data of smart meter test users in Taiwan from 2017 to 2018 and the public weather forecast data from Central Weather Bureau, trying to predict the electricity consumption of households in the next 24 and 168 hours. In the experiment, different models, input parameters and the time scale of input data are tried to get the best good prediction results. By accurately predicting the future electricity consumption, it can be provided to Taipower company as a basis for power dispatching and Time-of-use Rate pricing to enhance the integration and development of energy management and data science in Taiwan.