Federated personalized home BESS recommender system based on neural collaborative filtering

Home battery energy storage systems (HBESSs) has been experiencing an increasingly popularization and marketing process. This consequentially leads to an information filtering challenge for the residential customer to choose the most suitable HBESS products from the large number of candidates HBESSs in the market. This paper proposes a novel personalized HBESS recommender system to provide decision -making support for residential customers to make HBESS choice. The system makes HBESS recommendation following 2 stages: (1) in the first stage, the system uses a federated learning process to aggregately analyze the customers' preference tendencies on HBESS products from the datasets owned by different HBESS service providers without having the data actually exchanged; based on the learnt preference trends, the system generates a HBESS shortlist that are likely to fit the target customer's profile; and (2) in the second stage, the system further filters the shortlisted HBESSs by evaluating the household energy cost they can create for the target customer. Combining the considerations of both personal preference and energy cost, several HBESS products are finally selected from the previously generated shortlist and recommended to the target customer. Extensive simulations are conducted to validate the effectiveness of the proposed system.