Study On Multi-agent Evolutionary Algorithms For Mid-long Term Optimal Operation Of Cascade Hydropower System

China's booming economy inevitably leads to an increase in energy consumption.The increasingly serious environmental pollution is particularly remarkable.As a clean energy source,hydropower has the advantages of flexible dispatching and promoting the stability of power grid,and it plays a key role in the peak regulation and frequency modulation of power grid,is of great significance to the improvement of environment and economic progress.However,large scale cascade hydropower stations have many characteristics,such as multiple quantity of power stations,high power generation heads and large installed capacity.The midlong term optimal operation of cascade hydropower system is a typical large-scale multi-step problem with multi-variable,high dimension and complex constraints that may be difficult to solve.At present,there are many difficulties to be studied and solved.First,the dry season runoff forecast influencing scheduling of cascade hydropower system is studied.Then,based on the analysis of solving quality and efficiency of the mid-long term optimal operation of cascade hydropower system,the dissertation proposed two novel Multi-Agent evolutionary algorithms.The main research contents include:(1)On the basis of the analysis for the conventional formula of recession curve,the dissertation puts forward fitting recession coefficient by n-order polynomial function,builds improved recession curve forecasting models based on n-order polynomial,and selects the optimal model by comparing each average deviation rate of model.The reasonable forecast period is determined and the detailed forecast steps are presented.The dissertation takes the Gongguoqiao hydropower station as the example to conduct the daily runoff forecast for 30 days in dry season and takes the improved recession curve model based on 2-order polynomial corresponding to minimum average deviation rate as the optimal model.It can be shown in the forecast results that the proposed model can be available for reference to other power stations.(2)It proposed a Multi-Agent artificial fish swarm algorithm(MAAFSA)for solving the mid-long term maximum power generation problem of cascade hydropower system.Five Agent modules with different functions,including Core Agent,Group Agent,Action Agent,Evaluate Agent and Judge Agent,are constructed.The artificial fish Agent completes its iterative process by mutual collaboration among Agent modules and the self-learning of Group Agent,so that the convergence efficiency of artificial fish swarm algorithm(AFSA)is improved.Taking the four cascade hydropower stations in Wujiang River as an application example,the proposed algorithm is applied to solve the mid-long term maximum power generation problem,which verifies the effectiveness of the algorithm.(3)Combining Multi-Agent technology with Particle swarm optimization(PSO)algorithm,it proposed a new Multi-Agent particle swarm optimization(MAPSO)algorithm for solving the mid-long term maximum power generation problem of cascade hydropower system.MAPSO not only absorbs the PSO evolutionary mechanism where each Agent can share information with the global optimal Agent,but also has the ability to communicate and cooperate with the neighbors and self-learning ability.Therefore,MAPSO can effectively avoid local optimal behavior of traditional PSO,and greatly reduce computation time and efficiency,and converge to the global optimal solution more quickly.The method has been applied in the operation of hydropower stations in the Wujiang River.The results indicate that MAPSO has good calculation efficiency and optimization ability,which can be used as reference for other cascade hydropower stations.Finally,the contents of the full text are summarized,and the insufficient study and new research directions are prospected.