Research On Scenario Generation And Scheduling Of Multi-Source Power System Based On Finite Data Drive

Wind power output has obvious characteristics of randomness,fluctuation,and intermittence.With the increase of wind power permeability,the uncertainty of wind power output must be considered in power system scheduling.The traditional methods describe the stochastic and fluctuating characteristics of wind resources in different areas,but the wind resources in different areas are different,and the wind famm output characteristics are complex,so it is difficult to model separately for the geographical location of wind farms and the particularity of equipment.Power system scenario analysis is widely used because it can describe the uncertainty of wind power output probability.To solve these problems,this paper proposes a robust scheduling method based on the generation of wind power similar scene sets.Firstly,the characteristics of wind resources based on historical meteorological data of wind farms in different regions are studied from the aspects of wind power climbing day distribution and wind speed distribution.Secondly,a method of historical similar day selection based on weighted Euclidean distance and weighted correlation degree is proposed.The historical similar day of the day to be scheduled is determined by analyzing the daily numerical weather forecast and historical meteorological records Part constraint ensures that a similar day of a day is not less than 20 days,so as to build a training dataset for the day ahead scheduling scene generator.Then,for the similar daily historical output data,the self symmetric structure generator and discriminator are designed,and the Wasserstein distance is used to replace the JS divergence distance measurement adopted by the traditional Gan,to learn the intermal distribution law of historical data and generate the wind farm output scenario similar to the dispatching daily wind power output probability distribution.At the same time,by adding regular terms and optimizing the Lipschitz limit,the gradient penalty is applied to each similar day scene independently to avoid gradient disappearance and mode collapse.Then,a similar daily scenario reduction method based on heuristic synchronous generation is used to effectively reduce the generated scenario set and reduce the computation of the optimal scheduling scheme.Finally,based on the reduced scene set,the characteristics of wind power daily output are described.Considering the output constraints of wind power and conventional units,the objective function including the minimum operating cost of thermal power and the penalty function of wind abandonment is designed to take into account the economy and limit the air abandonment,to specify a reasonable day ahead generation plan and arrange coordinated generation of multiple power sources.Experimental results show that the typical wind power similar day scenario set with extreme scenarios generated by the algorithm in this paper is effectively reduced and then substituted into the unit combination model for the optimal solution.The application effect of the algorithm in tihis paper is verified by comparing the operation cost of the thermal power Unit,the average grid-connected power of wind power,and the system security in the day with extreme scenario scheduling.