Study Of Over-relaxed Dual-Q Automatic Power Generation Control Strategy Considering Carbon Emission

In order to accelerate the energy transition,control environmental pollution,and achieve the goal of "emission peak" and "carbon neutrality".Large-scale clean energy began to connect to the grid to alleviate the energy crisis and accelerate the completion of the "double carbon" goal of the power industry.However,wind power,photovoltaic and other new energy sources are random,intermittent and difficult to predict accurately,so large-scale grid connection will definitely lead to fluctuations in grid frequency,thus bringing serious challenges to the safety,stability and economic operation of the power system.Therefore,from the perspective of Automatic Generation Control(AGC),this paper explores a multi-intelligent cooperative control strategy and multi-layer power allocation strategy for distributed energy sources to achieve the optimal cooperation of distributed multi-area energy systems,and thus improve the control performance of the grid which is deteriorating due to the strong random disturbances brought by the large-scale grid integration of new energy sources.First,the current status of domestic and international research on automatic generation control is summarized and reviewed,and the power system AGC principle and system frequency modulation are elaborated.At the same time,the two types of interconnected grid AGC models built are introduced and analyzed.Second,for distributed multi-region interconnected power systems oriented to the integration of large-scale new energy sources,a strategy that can solve the strong random disturbances and frequency instability generated by new energy grid connection is proposed from the perspective of AGC.Considering the joint participation of battery storage and other conventional units in AGC frequency regulation,an over-relaxation double-Q algorithm(ORDQ)is proposed to solve the problem of difficult convergence when the intelligent body is caught in a self-loop by introducing a relaxation factor ω to accelerate the convergence speed of the algorithm,and to reduce the overestimation error of the value function by incorporating the double-Q strategy.The simulation of the real grid environment is carried out by building a two-area interconnected power system model and a multi-area interconnected power system model containing electric vehicles to verify the practical engineering effectiveness of the ORDQ algorithm.The simulation results show that the ORDQ algorithm has fast convergence characteristics and can solve the frequency instability problem and obtain the best control performance among the grid regions.Finally,a hybrid sampling two-layer power allocation strategy based on double-Q learning is proposed(HSDQ),which unifies the advantages of the off-and on-strategy in the reinforcement learning algorithm and improves the convergence accuracy of the algorithm by introducing the hybrid sampling parameter σ on the basis of double-Q learning.The HSDQ allocation strategy can realize the real-time dynamic allocation of AGC power commands to different generating units and achieve the total regulation power command The proposed algorithm is simulated and analyzed by using the carbon emission as the target reward function of HSDQ and in the improved IEEE two-region load frequency control model and the multi-area microgrid cluster model.The results show that the proposed allocation strategy has better control performance compared with other algorithms and can effectively reduce carbon emissions at the same time.