Table 1 Summary on EV charging & energy Management.

Zavvos et al⁹.​

Game-theoretic model with subgame-perfect equilibrium solution for competing charging stations

Models’ realistic investor and driver behavior; analyzes impact of subsidies on station operators and EV drivers

Limited to static analysis; assumes simplified cost structures

Gao et al.​¹⁰

Comprehensive demand response with cooperative game theory and real-time pricing strategy

Integrates demand response with cooperative game framework; optimizes distribution network charging;

Assumes cooperative behavior; may not reflect real-world competitive scenarios; limited validation on large-scale networks

Zhang et al.​¹¹

CDDPG (Charging control Deep Deterministic Policy Gradient) - deep reinforcement learning approach

Learns optimal charging control strategy in continuous action space; reduces charging cost while satisfying battery requirements;

Requires substantial training data; convergence speed depends on network architecture; may struggle with non-stationary environments

Park & Moon​¹²

Multi-agent deep reinforcement learning (MADRL) with centralized training and decentralized execution (CTDE)

Enables decentralized decision-making for multiple EVs; reduces system communication overhead; handles uncertainty in arrival/departure times;

Scalability concerns with very large EV fleets; complexity in coordination increases with agent count; requires careful reward function design

Bertolini et al.​¹³

Deep reinforcement learning for real-time power output optimization at smart charging hubs

Optimizes both profit and service quality simultaneously; handles uncertainty in EV arrivals; real-time operational capability

Limited field validation; assumes simplified hub architecture; does not account for battery degradation costs explicitly

Wang et al.¹⁴

Reinforcement learning (RL) with feature-based function approximation for joint pricing and scheduling

First model-free approach handling time-varying state/action spaces from random EV arrivals/departures;

Model-free approach may require extensive data; online learning convergence time not explicitly bounded; assumes simplified electricity pricing

Lee et al.¹⁵

Real-time intelligent energy management using reinforcement learning for hybrid electric vehicles

Handles complex powertrain dynamics; real-time decision-making for power split between ICE and electric motor;

Focuses on HEV rather than pure EV; integration with renewable sources (solar/wind) which adds complexity

Jia et al.¹⁶

Differential game theory model for dynamic pricing in EV market

Develops theoretical framework balancing pricing and demand dynamics; analyzes equilibrium pricing behavior;

Theoretical model with limited empirical validation; do not account for real-world market changes and transaction costs

Liu et al.​¹⁷

DDPG (Deep Deterministic Policy Gradient) reinforcement learning for dynamic pricing by EV aggregators

Handles continuous action space effectively; learns optimal pricing strategies for revenue maximization; model-free approach

Convergence to optimal policy not guaranteed; single aggregator assumption limits the applicability

Chavhan et al.​¹⁸

AI-empowered game theoretic model with deep neural networks and Nash equilibrium-based strategy

Combines neural networks with game theory; integrates decentralized renewable sources and battery storage; Hardware-in-Loop (HIL) validated;

Complex implementation requiring careful parameter tuning; computational overhead from dual optimization (neural network + game theory)

Piao et al.¹⁹

Non-cooperative game theory for decentralized charging station pricing

Simple decentralized approach without central coordination; encourages competitive pricing behavior

Limited to pricing decisions; does not optimize charging schedules; assumes perfect information about competitor strategies

Rasheed et al.​²⁰

Multi-leader multi-follower Stackelberg game with dynamic pricing and traffic flow integration

Addresses charging anxiety through real-time pricing and traffic information; minimizes total cost and time; Nash equilibrium ensures fairness;

Requires accurate traffic and pricing data; computational complexity of multi-level optimization

Muratori²¹

Empirical modeling of uncoordinated PEV charging impact on residential power demand

Highly resolved temporal and spatial analysis (10-minute intervals); identifies peak demand increases from uncoordinated charging

Primarily observational/empirical; does not propose control solutions; limited to residential charging scenarios;

Richardson²²

Literature review and modeling framework for EV grid integration and renewable energy coupling

Comprehensive review of modeling approaches; identifies key impacts on grid stability and cost; establishes framework for multi-scale analysis

Analytical review rather than novel solution; does not propose specific optimization algorithms

Ma et al.​²³

Decentralized charging control via Nash certainty equivalence principle with iterative negotiation

Decentralized implementation with low communication overhead; theoretically guaranteed convergence

Requires negotiation phase before charging; homogeneous PEV assumptions for perfect valley-filling; does not handle rapid demand changes

Dorokhova et al.​²⁴

Deep reinforcement learning with discrete, continuous, and parametrized action spaces for PV-integrated charging

Multiple algorithm formulations for different problem variants; improves PV self-consumption; outperforms rule-based and MPC benchmarks; real-world validated

separate models for different action space types reduce generalization; limited to single charging location

Heendeniya & Nespoli²⁵

Stochastic deep reinforcement learning agent for grid-friendly EV charging

Handles grid constraints explicitly; stochastic approach reduces peak demand impact; grid-friendly operation focus

stochastic formulation increases computational requirements; single station assumption limits grid-wide applicability

Fan et al.​²⁶

Improved MA-DDPG (Multi-Agent DDPG) for load frequency control in multi-microgrids with V2G

Coordinates multiple microgrids through multi-agent learning; balances EV discharge and MT costs; outperforms PID and traditional RL

Multi-microgrid setup increases complexity; hyperparameter tuning required for different grid configurations

Wen et al.²⁷

Deep Q-Learning for optimal microgrid scheduling with V2G capability

Online learning capability; handles uncertainty in renewable generation and EV behavior

Limited to discrete action spaces initially; convergence time dependent on state/action space size

Ke et al.²⁸

MPC-VSG (Model Predictive Control with Virtual Synchronous Generator) for frequency control in islanded microgrids

Improves frequency response dynamism; secondary frequency regulation capability; handles load and generation uncertainties

Requires accurate system model for MPC; VSG control adds complexity; focuses on frequency rather than complete energy optimization

Pan et al.​²⁹

Deep reinforcement learning for online V2G scheduling in multi-energy microgrids

Handles real-time optimization; multi-energy system integration; online learning without pre-optimization

Multi-energy modeling complexity; scalability to large microgrids untested

Zhang et al.³⁰

Deep reinforcement learning considering solar power forecast errors in V2G operation

Accounts for forecast uncertainty in real-world conditions; improves operational resilience

generalization to other renewable sources needs validation; battery aging from V2G not explicitly considered

Elkhodr et al.³¹

Semantic-aware intelligent V2G energy management framework

Semantic processing enables context-aware decision making; intelligent framework for diverse EV & grid requirements

Limited technical detail on semantic processing methodology; evaluation scope not clearly defined;

Shang et al.​³²

Spatio-temporal data fusion with Large Language Model (LLM) and LoRA for EV charging demand prediction

Novel LLM application in forecasting; multi-source data fusion (temporal, spatial, behavioral); real-world Beijing dataset (830 K + records)

LLM fine-tuning complexity; high computational overhead; requires extensive labeled data; forecasting-focused without optimization

Shang et al.³²

Non-cooperative game theory for V2V (Vehicle-to-Vehicle) charging mechanism

Accessible close-loop mechanism enabling peer-to-peer charging; reduces dependency on centralized stations;

Assumes perfect information between vehicles; peer-to-peer charging infrastructure requirements; assumes stationary vehicle locations