Research On Real-time Active Power Dispatch Model And Algorithm Of Wind Power Integrated System

Even though the great development of renewable energy like wind power can decrease the fossil energy consumption,release the environment pollution and be beneficial to sustainable development,its uncertainty and random fluctuation bring a lot of troublesome on the operation of the power system.Therefore,how to construct an accurate wind power output model,simulate changes in power flow and system frequency caused by the wind power fluctuation,and explore the dispatch model and its solution methodology which is suitable to the power system with wind power integration are of great significance.In order to solve the above issues,this paper aims to build a comprehensive real-time dispatch model considering the frequency adjustment characteristics of power systems and its solution methodology,focusing on multi-dimensional wind power output modeling,probabilistic load flow with frequency regulation characteristics considering correlated wind sources and the real-time active power dispatch with frequency regulation characteristics.This paper studies the probabilistic model construction,sampling and scenario reduction of multi-dimensional wind power.A dimension reduced Pair-copula model is adopted to describe the multi-dimensional wind power forecast error.Then the Latin Hypercube Sampling with Dependence(LHSD)and Affinity Propagation(AP)clustering method are used to obtain the samples and scenarios with suitable size.At first,Principal Component Analysis(PCA)or Local Preserving Projection(LPP)is adopted to map the original high-dimensional data into low-dimensional space,and model the low-dimensional data using Pair-copula theory.Then the forecast error samples and scenarios can be obtained by LHSD and AP clustering.Through modeling and simulation of real wind power data,it is verified that the reduced Pair-copula method can accurately describe the wind forecast error of multi-dimensional wind power and not only can capture their nonlinear dependence well,but also reduces the complexity of model construction.The sample generation and scenario reduction methods based on LHSD combined with AP clustering are very flexible and effective.Compared with the K-means method,the classification stability of AP is better and the obtained scenarios are more representative.This paper studies the Probabilistic Dynamic Load Flow(PDLF)considering the dependence among different wind sources,and an analytical algorithm based on Reduced Joint Moment Method(RJMM)is proposed.By adding system frequency as an unknown variable in the traditional power flow equations and the active power flow equation of the balancing node,the power-frequency regulation characteristics of generators can be fully taken into account,so that the power flow result is more in accordance with the actual situation.The case study shows that the power flow results obtained by PDLF is more in line with the actual operation condition,compared with the traditional probabilistic load flow.The proposed RJMM method can take into account the nonlinear dependence among different wind sources.At the same time,it can reduce the computation complexity by adopting LPP,and has high calculation accuracy and computational efficiency.This paper studies a chance constrained real-time active-power dispatch model based on PDLF,taking into account the effects of primary and secondary regulation of frequency caused by wind power fluctuation,so that the system has sufficient primary and secondary regulation capacity to cope with the random fluctuation of wind power.According to the case study,it is verified that the resulting dispatching results can deal with wind power fluctuation better and can help maintain the stability of frequency.In addition,RJMM is also used to determine the probability that the chance constraints are satisfied.The chance constrained model is transformed into a quadratic programming model,which not only can effectively account for the nonlinear dependence among different wind sources,making the dispatch results more accurate,but also simplify the solving procedure.Though the simulation results,RJMM is more efficient and accurate compared with other chance constrained programming solving methods,This paper studies an active real-time dispatch model based on AC dynamic power flow and its solution methodology.The real-time dispatch model is based on AC dynamic power flow and the reduced wind power scenarios.A homotopy continuation methodology is proposed to solve the non-convex optimization problem.In addition,in order to improve the solving efficiency,the Optimality Conditional Decomposition(OCD)is used to solve the non-convex optimization problem by parallel computation.According to the numerical simulation,it is verified that the proposed homotopy continuation method is robust in solving non-convex optimization problems.Compared with other methods,this method not only has fewer iterations and thus has higher computation efficiency,but also can overcome the drawbacks like being sensitive to the initial guess and having difficulties in dealing with ill-condition,thus it has better convergence property;Moreover,the proposed parallel computation method based on OCD can significantly reduce the computation time and has the potential to online application.