The operational needs of modern smart grids become more difficult to handle because their systems use renewable energy sources and people who produce energy for their own needs. Traditional demand response methods experience problems because they make decisions at slow speeds and their power reduction methods do not function effectively and their system users lack confidence in their remote system operators. The research team developed an Energy Efficient Hybrid LSTM–XGBoost Optimization system which works together with a blockchain-based demand response orchestration system to solve this problem. Long Short-Term Memory (LSTM) networks take demand elasticity and short-term consumption patterns from real-world data to create their predictions while XGBoost creates demand response solutions by analysing which loads should be reduced according to their user effects and energy savings potential. The permissioned blockchain smart contract system creates binding power to enforce demand response contracts while it checks compliance with responses and handles incentive payments without needing much processing power. The simulation results show that the system achieved 97.6% Demand Response Accuracy and 31.4% Peak Load Reduction Efficiency and 24.8% Energy Utilization Improvement while it decreased DR signalling time to 38 ms. The use of blockchain technology for coordination purposes results in 29.6% fewer response violation incidents. The proposed framework establishes a scalable, trust-aware, and energy-optimized demand response paradigm which next-generation smart grids can use.