Demand-Side Management of EV Integration with Lesco Grid Station

Abstract

With an estimated 500,000 EVs on Pakistani roads by 2030, Pakistan's National Electric Vehicle Policy (NEVP) 2019 aims for 30% of all new vehicle sales to be electric. This transformation is an acute threat to the infrastructure of Lahore Electric Supply Company (LESCO) which is already operating 146 grid stations having 10/13 MVA (132/11 kV) ONAN/ONAF power transformers with 80-90% loading of their ONAN capacity on peak summer evenings. At 30% penetration, the uncoordinated Level-2 EV charging causes the transformer to be loaded to 15.57 MVA (119.8% of its 13MVA ONAF transformer rating) for 2.5 hours per day, bringing the hot-spot temperature above the IEEE C57.91 rated limit of 110°C by 37.1°C and increasing the daily insulation loss-of-life by about 28 times compared with no EV charging. In this paper, a modular AI-Driven Demand-Side Management (AI-DSM) framework is proposed that addresses this crisis using only software - no hardware changes required. The framework consists of three successive steps: (1) an unsupervised identification of EV owners using k-means and DBSCAN clustering on load profiles of smart meters; (2) a 24 h-ahead transformer loading forecast based on a stacked Long Short-Term Memory (LSTM) neural network; (3) a quadratic-programme based charging-schedule optimiser that fills the overnight load valleys while taking into account the thermal capacity of transformers and the departure constraints of EV owners. The integrated framework has been validated against a 12-month simulation based on a calibrated LESCO Annual Report for 2022–23, and yields: EV identification F1 = 1.000 with zero false positives; load forecasting R² = 0.974, MAE = 0.168 MVA (53.8% improvement over seasonal naive baseline); peak load reduction 36.1% (transformer loading restored to 76.5% of ONAF rating); winding hot-spot held at 86.1°C (versus no-EV baseline of 85.2°C); and 100% user charging satisfaction — all without capital investment in hardware.