<p>Edge-enabled smart grid communication systems operate under tight energy budgets and fluctuating traffic conditions, making it essential to use routing and scheduling methods that control energy use without compromising delay or reliability. Many existing approaches treat routing and scheduling as separate tasks, which often leads to poor coordination—especially when traffic surges unexpectedly or when edge devices face uneven energy availability. In this work, a unified solution is developed that couples a Genetic Parser, used to generate energy-aware routing and slot assignments, with a deep residual model that leverages Bivariate Pascal statistics to anticipate short-term traffic and energy behavior. The Genetic Parser selects paths that distribute load efficiently, while the predictive module estimates the likelihood of delay increases or energy spikes so the scheduler can adjust in advance. Tests carried out on representative smart grid scenarios show clear gains: energy use drops by roughly 18–24%, network lifetime increases by about 27%, and end-to-end latency falls by 15–20% when compared with established methods. Packet delivery also remains consistently high, reaching 98.4% even when traffic conditions vary rapidly. These results indicate that combining evolutionary search with lightweight predictive modeling can improve both stability and overall efficiency in next-generation smart grid communication networks.</p>