Data-driven LSTM-HHO optimization framework for strategic V2G connection sitting

The large-scale integration of electric vehicles (EVs) introduces critical challenges in power systems, including increased power losses, voltage instability, and demand-side management complexity. This paper proposes a hybrid LSTM–Harris Hawks Optimization (LSTM–HHO) framework for coordinated vehicle-to-grid (V2G) planning and scheduling. The LSTM predicts EV charging priorities based on state of charge and parking duration over a 24-h horizon, while HHO optimally determines V2G placement and scheduling. A multi-objective formulation minimizes power losses, voltage deviation, and expected energy not supplied (EENS), while maximizing the loadability limit (LAL). EV uncertainties are modeled using Monte Carlo simulation. Validation on IEEE 9-, 26-, and 118-bus systems demonstrates that, compared to the no-V2G scenario, the proposed method achieves up to 21.1% reduction in power losses, 59.1% reduction in voltage deviation, 28.6% improvement in LAL, and 61.3% reduction in EENS, while consistently outperforming benchmark algorithms (ICA, MOGA, and GA).