考虑风电和负荷不确定性的输电网多目标柔性规划

风电大规模集中并网使输电网中电源出力的不确定性显著增强，并且当前考虑风电和负荷不确定性的输电网规划模型大多采用直流潮流，规划结果的准确性有待提高。以交流潮流为基础，构建了考虑风电和负荷不确定性的输电网多目标柔性规划双层模型。上层模型是以建设投资等年值费用、年网损费用和运行效率为目标的输电网多目标规划模型，采用带精英策略的非支配排序遗传算法求解，并将帕累托最优规划方案集合传递给下层；下层模型是以弃风和切负荷量最小为目标的多场景校验模型，用来校验帕累托最优规划方案集合对于风电和负荷不确定性的承受能力，根据下层模型的弃风和切负荷期望值修改上层模型的约束条件，从而将风电和负荷不确定性的影响传递给上层模型。对下层模型的交流潮流约束进行二阶锥松弛，将非线性且非凸的规划模型转化为具有凸可行域的二阶锥规划模型，利用Cplex求解器进行高效求解。最后，用调整后的39节点系统验证了所提模型的有效性。

Multi-objective Flexible Planning of Transmission Network Considering Wind Power and Load Uncertainties

Large-scale centralized integration of wind power into power grid significantly enhances the uncertainty of power output in transmission network. In addition, most of the current transmission network planning models adopt DC power flows, which affects the accuracy of planning results. Based on AC power flow, this paper constructs a two-layer model for multi-objective flexible planning of transmission networks considering the uncertainty of wind power and load. The upper-level model is a multi-objective planning model of transmission networks, with consideration of the objective of the annual cost of construction investment, annual network loss and operating efficiency. The non-dominated sorting genetic algorithm-II is used to solve the problem, and the Pareto optimal planning scheme set is transmitted to the lower level. The lower-level model is a multi-scenario checking model with the minimum wind curtailment and load shedding as objective, which is used to test the endurance capacity of the Pareto optimal planning scheme set to the uncertainty of wind power and load. The constraints of the upper model is modified according to the wind curtailment and load shedding expectations, thus transmitting the impact of wind power and load uncertainty to the upper-level model. The AC power flow constraints of the lower-level model is conducted for second order cone relaxation, and the nonlinear nonconvex planning model is transformed into a second order cone planning model with convex feasible region, which is effectively solved using the Cplex solver. Finally, the effectiveness of the proposed model is verified by an adjusted 39-node system.