Multi-objective optimal scheduling method of integrated energy system via Q-learning-based memetic algorithm considering gas boiler and power-to-gas

To address the inefficiencies in multi-objective optimization of current integrated energy systems (IES) involving heat, electricity, and gas, which struggle to balance economic efficiency, low carbon emissions, and energy utilization efficiency, a multi-objective memetic algorithm based on Q-learning (Q-MOMA) is proposed for system optimization. First, the structure of the heat-electricity-gas system is explained, along with how its energy interacts with the natural gas network, heat network, and upper-level power grid. The output powers of thermal power units, coal-fired power units, power-to-gas devices, heat storage tanks, and gas boilers are among the decision variables. The optimization objectives are the system's economic cost, carbon emissions, and energy usage efficiency. Then, the Q-learning algorithm adaptively adjusts the multi-objective memetic algorithm's critical parameters and solves the model. Finally, simulations based on actual data from a particular power grid are conducted under multiple scenarios, and comparisons are made with optimization algorithms such as NSGA-III, multi-objective particle swarm optimization (MOPSO), and multi-objective memetic algorithm to verify the effectiveness of the proposed multi-objective scheduling method. The simulation results show that the Pareto solution set obtained by the Q-MOMA algorithm is superior to those obtained by MOMA, NSGA-III, and MOPSO algorithms, with average improvements of 9.35 % in economic optimization, 28.05 % in carbon emission optimization, and 3.77 % in efficiency optimization. • An integrated thermal-electrical-gas system was built with P2G devices and boilers. • Optimization considers economy, carbon emissions, and energy efficiency objectives. • Combining Q-learning with a multi-objective memetic algorithm for optimizing IES.