Joint optimization of task offloading and energy trading in edge-enabled smart grids using deep reinforcement...

Area of Science:

• Smart Grid Technology
• Distributed Energy Resources (DERs)
• Mobile Edge Computing (MEC)

Background:

• Increasing DERs and IoT devices necessitate advanced computing in smart grids.
• MEC integration enables real-time data processing for prosumers.
• Cyber-physical coupling exists between computational offloading and P2P energy markets.

Purpose of the Study:

• To jointly optimize computational task offloading and P2P energy trading.
• To maximize long-term system utility considering throughput, latency, and economic incentives.
• To address challenges of dimensionality and stochasticity in edge-assisted smart grids.

Main Methods:

• Formulation as a mixed-integer nonlinear programming (MINLP) model.
• Development of a hybrid framework combining Deep Q-Networks (DQN) and a constraint-aware heuristic.
• DQN for adaptive offloading policies and heuristic for community energy balance.

Main Results:

• The proposed hybrid method improved average utility by 12.3%.
• Task delay was reduced by 16.5% compared to baseline strategies.
• Robust operational feasibility was maintained under strict constraints.

Conclusions:

• The hybrid DQN and heuristic approach effectively balances computational offloading and P2P energy trading.
• This method enhances smart grid efficiency and economic benefits for prosumers.
• The framework demonstrates a viable solution for complex edge-assisted smart grid ecosystems.