Online Electric Vehicle Charging Detection Based on Memory-based Transformer using Smart Meter Data

TL;DR

This work tackles online, unsupervised identification of EV charging events from streaming behind-the-meter smart meter data. It introduces a memory-based Transformer (M-TR) with a global memory encoder and a local memory decoder, plus dual memory compression to achieve linear-time inference. Anomaly scores are derived from reconstruction errors and thresholded dynamically with Streaming Peak-Over-Threshold (SPoT) using EVT-based thresholds. Across the Pecan Street dataset, M-TR demonstrates strong F1 and ROC performance while enabling real-time operation (~1.2 s per minute), illustrating practicality for DNOs and real-time grid management. The approach reduces reliance on labeled data and prior EV profiles, while providing robust online detection in diverse charging patterns.

Abstract

The growing popularity of Electric Vehicles (EVs) poses unique challenges for grid operators and infrastructure, which requires effectively managing these vehicles' integration into the grid. Identification of EVs charging is essential to electricity Distribution Network Operators (DNOs) for better planning and managing the distribution grid. One critical aspect is the ability to accurately identify the presence of EV charging in the grid. EV charging identification using smart meter readings obtained from behind-the-meter devices is a challenging task that enables effective managing the integration of EVs into the existing power grid. Different from the existing supervised models that require addressing the imbalance problem caused by EVs and non-EVs data, we propose a novel unsupervised memory-based transformer (M-TR) that can run in real-time (online) to detect EVs charging from a streaming smart meter. It dynamically leverages coarse-scale historical information using an M-TR encoder from an extended global temporal window, in conjunction with an M-TR decoder that concentrates on a limited time frame, local window, aiming to capture the fine-scale characteristics of the smart meter data. The M-TR is based on an anomaly detection technique that does not require any prior knowledge about EVs charging profiles, nor it does only require real power consumption data of non-EV users. In addition, the proposed model leverages the power of transfer learning. The M-TR is compared with different state-of-the-art methods and performs better than other unsupervised learning models. The model can run with an excellent execution time of 1.2 sec. for 1-minute smart recordings.

Figures (7)

• Figure 1: Overview of the proposed M-TR for online EV charging identification. Given a live-streaming smart meter readings, M-TR sequentially identifies the EV charging happening at each sampling time $T$ without accessing any future reading. The dashed blue indicates the data flow of memories using the first-in-first-out (FIFO) principle.
• Figure 2: Overview of Memory-based Transformer (M-TR) Anomaly Detection (M-TR AD). The network consists of a memory-based transformer (M-TR) encoder and decoder units. M-TR encoder encodes the global memory smart meter reading $gm$ into long memory embedding of size $e_1$ of latent features. The M-TR decoder leverages contextual information from the compressed global memory with the local memory $lm$ for anomaly detection. By referencing relevant context information, the M-TR decoder enhances its ability to detect anomalies in the streamed data. The M-TR encoder and decoder units are based on the Transformer decoder unit in vaswani2017attention. During the inference phase, M-TR takes every incoming smart meter reading in an online fashion with the absence of future readings.
• Figure 3: Overview of the Transformer decoder unitvaswani2017attention used in our M-TR anomaly detection network.
• Figure 4: Examples of the smart meter recording in the dataset used in our experiments and their anomaly distribution (EV-charging) as shown in red color.
• Figure 5: Snapshot of the smart meter recordings in the dataset used in our experiments and their temporal annotations of EV charging label as shown in the red color.