Online Operation of Renewable Energy and Battery Integrated Electric Vehicle Parking Lots with an Improved Priority Rule

Deploying renewable energy generation and battery storage can reduce the operation cost of electric vehicle parking lots. Given uncertainties from vehicle arrival times, charging demands and renewable output, operating a parking lot without exact forecasts of uncertainty is a challenging problem. This paper proposes an improved priority rule based on vehicle charging deadlines and remaining energy needs. When the charging power is continuously adjustable, the proposed priority rule outperforms the state-of-the-art charging policies. It is proven that there exists an optimal charging policy that follows this priority rule. With the improved priority rule, it is only necessary to determine the total charging demand and battery dispatch, significantly reducing the decision space. This simplification allows the experience-driven K-nearest neighbor (KNN) method to effectively estimate charging and dispatch decisions, mitigating uncertainties without extensive historical data. The total charging power is then allocated according to the proposed priority rule. We further suggest a simulation procedure to evaluate the cost-to-go (remaining cost from the current to the final timeslot) given the currently observed state and some promising candidate actions. This could improve the prediction-free KNN policy. Experimental results on real-world data show that the proposed KNN-integrated priority rule outperforms existing charging strategies, achieving cost reductions of up to 10%, with online simulation further boosting performance by an additional 10%.