EV charging load forecast using LSTM: a case study at a multi-functional building in Uppsala, Sweden

Energy-smart facilities contribute to a cleaner future by assisting the decarbonization of power generation and transportation. One example is the multi-functional building called Dansmästaren, which comprises a residential part, a commercial part and a parking garage and that is located in Uppsala, Sweden. This facility is a test bed for research on smart infrastructures for electric vehicle charging and has the flexibility to integrate different energy storage technologies, electricity generation units, database collection, and connectivity systems. Furthermore, Dansmästaren’s parking garage targets two distinct EV owner groups: public charger users and those with residential charging access. This work provides insights into charging behavior across these categories. This paper applies load forecast model using LSTM neural network to predict the parking garage’s load demand profile at Dansmästaren based on historical data. The performance of the LSTM network during testing and training is analyzed across various scenarios, using historical data from May 2021 to April 2023 collected at the parking garage. The hour-ahead load demand prediction achieves a mean absolute error (MAE) of 4.7. The results obtained in this study are valuable for implementing future smart charge strategies in real environment and for increasing knowledge about EV charging patterns.