| Hydropower is an important renewable energy in China.Hydropower stations need to meet the shipping,irrigation,water supply and other ecological needs while ensuring the maximization of their own benefits.These needs are often reflected in the optimal scheduling model as soft constraints.Due to the high complexity of hydropower basin background,and external uncertainty factors such as random inflow have great influence on power station decision.Therefore,how to solve the priority problem of long-term constraints of hydropower stations under random conditions has become the key to be solved.In this paper,two refined MLDC models are proposed based on the multi-level dependent chance(MLDC)reservoir scheduling background for solving the soft constraints in the optimization model of single-stage hydropower stations and cascade hydropower stations.The specific research contents are as follows:(1)Model 1 MLDC-RC is a MLDC model based on reliability opportunity dynamic programming(RCDP).In the solution process,all soft constraints are replaced by the opportunity constraints of 1,and the priority of constraints is reflected by setting Lagrange multipliers of different levels of constraints.Based on the probability distribution random inflow and probability distribution random inflow,the model is refined into MLDC-RC-SDP and MLDC-RC-SSDP.Multi-optimization severely limits the accuracy of stochastic dynamic programming(SDP)in solving non-convex maximization reservoir operation model.In order to solve this problem,a two-stage optimization algorithm combined with traversal and search is introduced.Through coarse traversal of the whole feasible region,single or multiple feasible regions that may have local optimal solutions are identified,and local search algorithm is used to optimize each feasible region.The model and method are applied to Xiaowan Reservoir and cascade reservoirs in an example,and the results are analyzed.(2)Model 2 MLDC-DC is a MLDC model based on correlated chance dynamic programming(DCDP).The constraint priority of the model can be explicitly expressed in the recursive function of dynamic programming(DP).In DCDP model,all soft constraints and benefits of reservoirs are defined as objectives with different priorities,and objective vectors are used instead of objective functions in general dynamic programming.Replace the objective maximization logic of each state solution with the comparison logic of higher priority and better performance.Combined with the two-stage optimization algorithm described in the previous chapter,MLDC-RC-SDP,MLDC-RC-SSDP,MLDC-DC-SDP and MLDC-DC-SSDP are used to solve the multi-layer constrained reservoir scheduling model.The minimum discharge flow,minimum output and guaranteed output are taken as the three-level constraints,and the maximum power generation is taken as the objective function.The results are analyzed.Case studies show that the proposed hydropower station long-term constraint classification model based on correlation chance programming can solve the soft constraint classification problem of hydropower dispatching model.Among them,the MLDC-DC-SSDP model has better results than the other three models,which can more fully reflect the constraint priority and provide corresponding basis for scientific and reasonable formulation of operation scheme. |
