Research On Low-carbon Economic Dispatch Of Carbon Capture Power Plants Under Uncertain Environmen

In order to achieve the strategic goal of carbon neutrality,how to reduce the carbon emissions in the process of power generation has become a must to be considered in power grid scheduling.At present,thermal power with stable output power is always an important source of electric energy in the existing power system.On this basis,the use of renewable energy such as wind power can achieve low carbon emissions.However,renewable energy is affected by seasons and regions easily,so the power system will face great uncertainties after the addition of renewable energy.In addition,in some regions,coal-fired power plants will be transformed into carbon capture power plants,which has more equipment limitations,which also brings new challenges to the design of optimal economic scheduling algorithm.Economic dispatching problem refers to the problem of adjusting the output power of units to minimize the cost under the constraints of a large number of power systems.Therefore,when the power system is more complex,it also brings new challenges to design the optimal economic scheduling algorithm.Besides,because the safe operation of power system needs to meet many constraints,the scale of economic dispatching problem is generally large,and the actual calculation cost is high.Today,most methods are one-sided,either failing to take into account the problem of reducing carbon emissions,or are expensive to calculate.Computer technology is good at processing massive data and provides a new solution to the problem of fast economic dispatching.In this paper,considering the combined power system of carbon capture power plant and wind farm in the scenario of wind power,a framework is designed to reduce the operating cost and carbon emissions of the power system,which provides a method to achieve the strategic goal of carbon neutrality.The randomness of wind power brings uncertainty to the power system,and the accompanying constraints in the operation process also bring difficulties to the calculation of economic dispatch.Therefore,in the uncertain scenario of power system,this paper proposes the method of combining reinforcement learning and deep learning to reduce the complexity of economic dispatching problem as much as possible,and finally improve the computing speed.The details are as follows:Firstly,an algorithm based on deep reinforcement learning is designed to obtain the optimal economic scheduling strategy,and the idea similar to pre-training is used in the framework proposed.The basic structure and weight of reinforcement learning are provided through initial-training,and the weight is fine-tuned in the later training to improve the efficiency of training.Secondly,in order to further reduce the computing cost,clustering is used for scene identification,deep neural network and perception,and the relationship between the required load and the constraint conditions of the power system is fitted,thus forming preliminary scheduling results,significantly reducing the scale of the proposed problem,and rapidly obtaining the optimal scheduling strategy.Thirdly,the strategy of economic scheduling is optimized by deep reinforcement learning,and the weight in the training process is fine-tuned to improve the actual calculation speed.Finally,by substituting the actual data into the framework for calculation,the experimental results show that compared with the existing technology,the proposed framework can provide an effective scheduling strategy,and the obtained results violate fewer constraints and less time than those obtained by traditional solvers.