Data Driven Economic Dispatch With Large Scale Wind Power Integrated Into The Power System

Large scale wind power integrated into the power grid can reduce the fossile fule depletion and can reduce the pollution to the environment.However,the uncertainty and volatility of wind power bring significant challenge to power system secure and economic dispatch.In this background,there is a significant meaning to conduct research on large scale wind power integrated power system economic dispatch.The current large scale wind power integrated power system economic dispatch relies on accurate wind power probability function.However,it is hard to obtain the accurate wind power probability distribution in reality.In most situation,we can only get statistical information of wind power.Thus,it is urgent need to research intraday and dayahead economic dispatch model and considers the effect of complete data and incomplete data on this model.For intraday economic dispatch,when there is complete data,this paper researches the data driven robust intraday economic dispatch and reserve determination model to overcome the overconservative issue of current intraday economic dispatch model.When there is incomplete data,this paper researches the distributionally robust intraday economic dispatch and reserve determination model with conditional value at risk recourse function to avoid the highly load shedding and wind spillage risk.For day ahead economic dispatch model,under the complete data condition,this paper researches the security constrained multistage robust dayahead dynamic economic dispatch with storage to solve the multi uncertainties and multi flexible resource coordination optimization problem.This paper aims at addressing the above problem from the four aspects:1.Aiming at addressing the current uncertainty sets have not fully captured the correlation of wind power and contain the scenario that rarely happen but will makes the optimization results overconservative,this paper proposes data driven uncertainty set modeling method for large scale wind power.This method uses nonparametric DPGMM model for wind power forecast error,and then uses variance inference algorithm to extract the GMM parameter information embedded within historical data.Based on these parameters information,we proposed a data driven polyhedral uncertainty set.The simulation results indicated that the proposed uncertainty set had a smaller volume than other data driven uncertainty sets with the same predefined coverage rate.This feature is more obvious for large scale wind power.2.Aiming at addressing the overconservaive issue of the current intraday economic dispatch which has not fully used the distribution parameters information from historical data when there is complete data,this paper proposes data driven intraday robust economic dispatch and reserve determination method.This method uses the proposed data driven uncertainty set to represent wind power uncertainty.This paper employs master and sub problem framewrok and uses column and constraint generation method to solve the proposed data driven optimization method.The master problem determines the generator output,up and down reserve.The subproblem conducts the robust feasibility check test.The simulation results demonstrated that the proposed optimization method is obvious better than traditional data driven robust optimization methods and distributionally robust optimization methods in terms of operation cost and reserve cost.3.Aiming at addressing the high load shedding and wind spillage of current intraday economic dispatch and determination model under the incomplete data condition,this paper proposes a distributionally robust economic dispatch and reserve determination model with conditional value at risk(DRDER-CVa R)recourse function to control the intrady operation risk,where the CVa R recourse function is used to measure the risk of load shedding and wind spillage.The DRDER-CVa R is equivalent to a semidefinite programming(SDP).The delayed constraint generation(DCG)algorithm and the alternate convex search(ACS)algorithm are employed to solve this model.The proposed DRDER-CVa R is compared with the existing economic dispatch and reserve determination(DER)model.The simulation results demonstrate that the DRDER-CVa R model can effectively control the risk of load shedding and wind curtailment according to the risk preference of the operators.4.Aiming at addressing the multiple uncertainties and uncertainty evolution in the dayahead economic dispatch,this paper proposes a security constrained multi-stage robust dynamic dayahead economic dispatch model with storage(SMRDEDS)to address the multiple uncertainties.Aiming at addressing the convergence issue of SDDP and computation efficiency issue when considering the N-1 constraints in multistage framework,this paper proposes a combined two stage benders decomposition and robust dual dynamic programming(RDDP)to solve the proposed model.This algorithm first uses benders decomposition algorithm to relax SMRDEDS model to a master and sub-problem.Then a robust dual dynamic programing(RDDP)algorithm is proposed to solve the multistage framework.Compared with SDDP algorithm,the proposed RDDP algorithm uses inner approximation and outer approximation methods to approximate the upper and lower bound of future cost-to-go function.Then this method uses backward pass and forward pass to update the upper bound and lower bound of the future cost-to-go function.When upper bound approximate lower bound,the algorithm terminates.The simulation results indicate that comparing with adaptive robust dynamic economic dispatch model,the SMRDEDS model has less operation cost and can provide a global optimal solution.Comparing with multistage stochastic dynamic economic dispatch,the SMRDEDS model can achieve deterministic convergence.Comparing with traditional deterministic contigency reserve,this model uses storage and generator to provide reserve according to the operation condition.Thus,this model can provide less reserve and reserve cost.