Research On Application Of Improved Particle Swarm Algorithms In Stochastic Optimal Dispatch Of Hydrothermal Power System

With the expanding scale of socio-economic,the crisis and issues that humanity has been facing are becoming increasingly prominent.Water has the advantages of abundant,zero pollution,renewable and high capacity.Therefore,hydropower has become an important form of development and utilization water,which not only providing cheap energy for social development,but also easing global warming,environmental degradation and other pressures.The scale of hydropower infrastructure has expanded rapidly and formed many systems of cascade hydropower stations,which constitute the most important part of electric power system along with the thermal power.However,the thoughts of optimal management of hydropower is not popular yet,which resulting in water can not be fully utilized.Based on probability theory,optimization theory and neural network theory,this paper analyzed the optimization scheduling problem of joint operation between the fire force units and cascade hydropower stations in flood,formating a stochastic optimal scheduling method of hydrothermal based on improved BP neural network PSO,which provides a theoretical basis for optimal operation of hydro-thermal jointFirstly,in order to analysing the random nature of single hydropower runoff,Pehrson-? distribution is used to describe the probabilistic characteristics of single hydropower runoff;then considering the Independent characteristics of each other among cascade hydropower runoff,this paper depict the probability correlation under the tail among the cascade hydropower runoff by Clayton-Copula function from Copula theory.Studies shows that Pehrson-? distribution function can fit the probability distribution of single Hydropower Runoff accurately;In addition,The joint probability distribution model of cascade hydropower runoff constructed by Clayton-Copula function,which compared with copula function without considering the relevance and other types of mutual,and having the highest goodness of fit,is the most accurate.Secondly,considering the cost of ultra-cost storage losses of cascade hydropower stations in flood season and the operating costs of thermal power generating units,a stochastic optimization scheduling model is structured.then,considering the randomness of runoff,the model is converted to deterministic optimization model using the average approximation by introducing chance-constrained programming theory.Further,in order to improve the computational efficiency of stochastic optimization scheduling model of hydrothermal,a improved particle swarm optimization is proposed based on BP neural network,with good nonlinear mapping ability and learning ability of BP neural network,and to quantitative modeling the Population evolution of Particle Swarm Optimization,which avoiding the problems of local optimum or slow convergence because of the unreasonable parameter of standard particle swarm optimization.Examples show that the improved PSO based on BP neural network can quickly get global optimal solution compared with the standard PSO in solving stochastic optimization scheduling problem of hydrothermal...