An LSTM-based Electricity Consumption Model for Load Aggregator Under User Response Mechanism

The load aggregator (LA) establishes its own electricity consumption model, enabling it to effectively grasp its ability to respond to the power grid. The large number of internal users in LA and the randomness of user responses pose great challenges in establishing an overall electricity consumption model. Therefore, for LA containing multiple types of users, in this paper we propose a method for building an LA electricity consumption model based on long short-term memory (LSTM) networks. Firstly, based on the response characteristics of internal users in LA, users are classified according to their incentive methods, and user characteristic data such as daily temperature, light intensity, incentive prices for each user, and user load baseline are aggregated to generate training samples. Subsequently, the LSTM network is trained based on the aggregated training samples and a mapping relationship between LA feature data and its electricity consumption curve is established. Finally, a case study of an LA with four types of users, including residents, commercial buildings, charging stations, and hospitals, is conducted for verification. The results demonstrate that the model can effectively characterize the electricity consumption behavior of LA after implementing user demand response (UDR).