Research On Robust Optimal Scheduling Of Active Distribution Network Considering Source-load Uncertainties

Active distribution network is an important technical means to accommodate renewable energy and improve the quality of power supply on the user side.However,with the increasing penetration of renewable energy,increasing load types and increasing load demand in active distribution network,the uncertainty of renewable energy and load poses a challenge to the safty and stability of active distribution network.On the other hand,the controllable resources connected to active distribution network are also enriched year by year.Although the control mode is more complex than traditional distribution network,it can further improve the capacity potential of active distribution network for renewable energy,and solve the disadvantages of the photovoltaic and load uncertainty to ensure the efficiency,economics,safty and stability of active distribution network.Therefore,in this paper,the uncertain optimal dispatch strategy of active distribution network is studied.Firstly,the basic theory of robust optimization model is presented,and its application in active distribution network is analyzed.The mathematical form of robust optimization and common uncertain sets are introduced,and three solving algorithms are introduced and compared on the basis of two-stage robust optimization.Further,the subproblems of C&CG algorithm and Benders algorithm,which can accurately solve the two-stage robust optimization model,are analyzed,and then the limitations of the two algorithms on large-scale uncertain variables are pointed out.In addition,for the multi-time scales robust optimization of active distribution network,the C&CG algorithm can be used when only the uncertainty of renewable energy is considered.However,it is difficult to apply the C&CG algorithm when considering the uncertainty of source and load.The second-order cone programming-linear power flow method proposed is used to solve the problem.Secondly,the interval prediction techniques of photovoltaic outp.u.t and load power based on the Gaussian mixture model are studied.The long short-term memory neural network is used to predict the photovoltaic power under three weather conditions as well as three types of loads power.Photovoltaic and load are divided into A type and B type according to the the prediction obtained errors,respectively.Further,the expectation maximization algorithm is used to solve the parameters of Gaussian mixture model,and the interval prediction of photovoltaic and load is accomplished at a certain confidence level.Simulations of photovoltaic and load forecasting in a certain geographic area show that the proposed model has some advantages over other single distributions,and can also guarantee reasonable coverage probability and normalized average width.Thirdly,a robust optimal economic dispatch strategy with multi-time scales considering the uncertainty of photovoltaic outp.u.t is studied.The controllable resources in the active distribution network are modeled,including capacitor banks,interruptable loads,energy storage system,distributed generators,static reactive power compensation device and distributed photovoltaic.As for the power flow constraints in the active distribution network,the second-order cone programming is used to relax it convexly.Based on the interval prediction of photovoltaic,a two-stage robust optimization model with multi-time scales considering uncertainties of photovoltaic outp.u.t is constructed on day-ahead scale and intra-day time scale.In the model,slow control equipment such as capacitor banks is determined in the day-ahead scale optimization,and fast control equipment such as distributed generators is determined in the intra-day time scale optimization.An example of an improved IEEE33 system is shown to verify the robustness of the optimal decision made by the proposed model,which ensures the system to operate safely and steadily within the fluctuation of photovoltaic and load.Finally,a fast solution strategy for multi-time scales robust optimization considering source-load uncertainty is studied.The linear power flow algorithm adapted to active distribution network is derived,and the relationship between node injection power and node voltage is obtained.Based on the linear flow,node voltage correction constraints and line current correction constraints are added to the original deterministic second-order cone programming relaxed power flow model,and a robust optimization model based on linear power flow and second-order cone programming is constructed.The robust optimization model is transformed equally using Karush-Kuhn-Tucker condition.The improved IEEE33 node verifies that the optimal decision made by the proposed model can effectively alleviate the adverse impact of adverse scenarios on the security and stability of active distribution network.