Yu Yang

As the number of electric vehicles (EVs) continues to grow, the demand for charging stations is also increasing, leading to challenges such as long wait times and insufficient infrastructure. High-precision forecasting of EV charging demand is crucial for efficient station management, to address some of these challenges. This paper presents an approach to predict the charging demand at 15-minute intervals for the day ahead using a multivariate long short-term memory (LSTM) network with an attention mechanism. Additionally, the model leverages explainable AI techniques to evaluate the influence of various factors on the predictions, including weather conditions, day of the week, month, and any holiday. SHapley Additive exPlanations (SHAP) are used to quantify the contribution of each feature to the final forecast, providing deeper insights into how these factors affect prediction accuracy. As a result, the framework offers enhanced decision-making for infrastructure planning. The efficacy of the proposed method is demonstrated by simulations using the test data collected from the EV charging stations at California State University, Long Beach.

This paper presents a framework for the optimal design of a solar and battery assisted electric vehicle (EV) charging station in southern California, with a focus on maximizing long-term profits while addressing operational uncertainties. The problem is conceptualized as an iterative two-stage decision process. In Stage I, the sampled designs of station infrastructure, including the number of chargers, the size of photovoltaic (PV) array, and capacity of the battery energy storage system (BESS), are specified. In Stage II, the EV charging rule is designed and simulated based on the charging request and solar power datasets. A model predictive controller and an empirical rule-based approach with incoming car forecasts are developed and compared for the vehicle charging management. The simulated annual operational profit and infrastructure investment with the consideration of long-term battery degradation is synthesized to build response surface for better design exploration in Stage I. Our results show that the proposed rule-based approach is computationally more efficient and suitable to integrate with response surface methodology (RSM) for design optimization. In addition, RSM is compared with adaptive particle swarm optimization (PSO) with multiple trials to demonstrate its superiority in high-profit design.

This article presents a robust model predictive control (R-MPC)-based scheduling approach for electric vehicle (EV) charging service at a station with power supply from a photovoltaic (PV) system and grid. The designed algorithm aims to maximize the operational revenue while handling high power loads in the grid, intermittent solar power supply, charging demands uncertainties, and online computational burdens. To this end, the scheduling problem is formulated as a mixed-integer nonlinear program (MINLP) with solar power generation and charging demands models embedded. The time-of-use price is employed for cost calculation. The resulting optimization problem then can be further simplified as a mixed-integer linear program (MILP). The proposed R-MPC is evaluated in the IEEE-13 node test feeder with simulated power load profiles and real charging demands data for demonstrations. The results show that R-MPC achieves near-optimal profit and meets all charging requests even under high uncertainties.

This article presents a cooperative volt-ampere reactive (C-VAR) control strategy to regulate voltage while reducing incurred power loss for a photovoltaic (PV) enabled grid experiencing abnormal (unexpected and extremely high) real power loads. The proposed C-VAR method only requires limited communications about power flow and voltage in a subset of nodes locally and is optimization-free. Thus, it is particularly suitable for the real-time voltage control for a distributed large-scale power grid. For demonstration purposes, C-VAR is evaluated using the IEEE 123-node test feeder and compared with benchmark solutions, such as the IEEE Standard 1547 and centralized model predictive control. The results show that C-VAR does not incur too much power loss and quickly drives the voltage back to the desired range under high power loads and PV indeterminacy.

This paper presents the feature extraction, selection and K-Nearest Neighbors (K-NN) algorithm to classify behaviors of sharks based on the data collected by tri-axial acceleration data loggers (ADLs). Because these behaviors are hard to observe in the wild and do not occur frequently, being able to adequately classify them is extremely challenging. In the proposed scheme, we first employ several transformations to enrich the static and dynamic acceleration data. Then, the enhanced data is converted from time to the frequency domain through the fast Fourier transform (FFT) for noise removal. A modified K-NN approach integrated with feature selection is developed and conducted on the frequency domain data to improve the F1-score for minority classes. The acceleration data of California horn sharks (Heterodontus francisci) gathered through ADLs mounted on the first dorsal fin is used to demonstrate our algorithm. A comparison study shows that the features extracted and selected by our proposed scheme can significantly improve the performance of classification on the imbalanced dataset.