Experiences in Predicting Port Availability
In the ever-evolving landscape of electric vehicle (EV) infrastructure, predicting port availability is a crucial component in alleviating range anxiety for EV users. A recent study has delved into this challenge, employing a rigorous evaluation framework to test the efficacy of predictive models in real-world scenarios.
Evaluation Methodology
Our evaluation was meticulously designed to reflect real-world conditions. For both 30 and 60-minute time horizons, we assessed forecasts at 100 randomly chosen stations. Each station’s occupancy status was sampled 48 times daily, at 30-minute intervals, over the course of an entire week. This comprehensive approach ensured that the data collected was both extensive and representative.
A Challenging Baseline
The predictive model’s performance was compared against a robust baseline known as the “Keep Current State” approach. This baseline operates on the premise that the number of ports available in the future remains the same as the current number. While simplistic, this approach is notoriously difficult to surpass, particularly in short-term predictions.
For instance, data from the East Coast of the United States indicated that less than 10% of ports change their availability status within a 30-minute window. Consequently, the assumption that availability remains constant proves accurate most of the time, posing a significant challenge for models attempting to add predictive value.
Key Metrics for Success
To gauge the model’s accuracy, we focused on two primary metrics: mean square error (MSE) and mean absolute error (MAE). The ratio of MSE/MAE to at least one free port was crucial in evaluating the model’s effectiveness in answering the critical binary question for users: “Will I find at least one free port (Yes/No)?”
This metric is vital as it directly impacts user experiences and trust in EV infrastructure. Accurate predictions can significantly enhance the user experience by reducing uncertainty and improving the efficiency of EV charging station usage.
For more in-depth information and data, you can explore the full study Here.
“`
