Probabilistic assessment method for evaluation of adjustable capacity of electric vehicle charging stations for volt–var control in distribution networks

The topology of charging pile converters is similar to that of static var. compensators. Theoretically, reactive power compensation can be achieved by improving the topology and control strategies, which enable voltage regulation in distribution networks (DNs). Technologies utilizing the adjustable capacity of electric vehicle charging stations (AC-EVCS) for Volt–Var control have recently attracted considerable research attention for accurate assessment of AC-EVCS capability. However, several persisting challenges hinder the accurate assessment of AC-EVCS capability. To address these limitations, in this study, we proposed a probabilistic assessment method for AC-EVCS in DNs. First, a correlation analysis method for factors that influence AC-EVCS was developed. AC-EVCS characteristics were defined, and the influencing factors were analyzed based on the statistical theory. Temperature, season, weather, and day type were identified as key factors. Second, we formulated a similar-day (SD) clustering method that considered the AC-EVCS influencing factors. In this method, the constructed influencing-factor-feature matrices of all historical days were clustered into multiple SD datasets using the deep convolutional embedding clustering model to obtain a robust data foundation for AC-EVCS assessment. Third, we proposed an AC-EVCS probabilistic assessment based on an improved L-transformer model and featuring an improved joint loss function that incorporated capacity constraint penalties and accounted for data fluctuation impact on assessment performance. This method demonstrated enhanced result accuracies, improved precision, and increased model robustness. Finally, validation using actual operational data from a southern Chinese region confirmed the assessment accuracy and practical applicability of the proposed method.