A Thermal Energy Usage Prediction Method for Electric Thermal Storage Heaters Based on Deep Learning

摘要

The electric thermal storage heater is a new type of electronic equipment which can provide full day heating by making use of low-priced electricity provided by power plants during nighttime. However, the traditional electric thermal storage heater cannot make use of the relationship between the thermal storage time and the thermal demand of each user, which may cause energy waste when the thermal energy stored exceeds the user's demand or thermal energy shortage caused by insufficient reserve. Aiming to solve the problems, this paper presents a method for thermal energy usage prediction based on recurrent neural networks. Considering the continuous variability of environmental data, we use Bi-directional LSTM to achieve more accurate prediction results by combining the correlation between past information and future information. In addition, aiming at the problem of weight distribution, the attention mechanism is used the prediction algorithm and achieve the goal of rationally assigning weight parameters which can effectively reduce the error rate after the model is updated. Experiments show that the proposed algorithm with self-learning ability can reasonably predict the user's thermal storage time. It can effectively avoid the energy loss caused by storing too much thermal energy while satisfying the daily thermal energy needs of users.