Battery capacities of EVs have reach up to 100kWh level these days. Fulfill such high power needs of EVs can be costly, EV users and fleet managers have become more serious and sensitive to the fluctuation of power price. To realize economical EV charging scheduling in the context of dynamic price electricity market, the forecasting of electricity price is of crucial importance. This research introduces a method to predict next-day electricity prices to 5-minute level, based on a model combined with 8 pieces of Artificial Neural Networks (ANN). Each ANN has one hidden layer with 20 neurons. The combined ANN model is then used to predict power price or Time-of-Use (TOU) price for the next day, with 5-minute accuracy. The input of ANN is simply the next day’s detailed 24-hour timestamp, in UNIX format. The predicted price results are used to establish the reward for EV scheduling actions on each time block next day. The reward matrix can be further used to solve the scheduling problem with q-learning framework. A detailed explanation of the training of the models and price forecasting results are presented.
EV Charging Management with ANN-Based Electricity Price Forecasting
2020-06-01
6628837 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Intelligent charging control method and system based on floating electricity price
Europäisches Patentamt | 2023
|Charging control method based on real-time electricity price, intelligent socket and charging pile
Europäisches Patentamt | 2021
|Battery exchange station charging scheduling method based on time-varying electricity price
Europäisches Patentamt | 2024
|Europäisches Patentamt | 2023
|