Research And Application Of Firefly Algorithm Based On Multi-strategy And Level-Based Learning

Firefly algorithm is a bionic optimization technology,which has the advantages of few parameters and easy to implement.It is widely used in science and engineering fields.However,the fitness difference between the particles in the population is ignored in the evolution process,which leads to the problems that the algorithm is prone to fall into local optimal and low optimization accuracy.Level-based thought originated from animal populations phenomenon existing in the class,in nature,biological systems showed obvious characteristic and dynamic characteristic of social class for a real biological communities,populations can be divided into several levels,at different levels of individuals usually plays a different role,each individual can according to their different levels of play the role of the most appropriate.Aiming at the problems existing in firefly algorithm,this paper introduces the idea of level-based and divides the population into several levels.Water resource is a kind of safe,efficient and renewable natural resource.Rational use of water resource can effectively alleviate the problem of energy shortage.Runoff forecasting and reservoir dispatching are two important technologies for rational utilization of water resources.How to improve the performance of these two technologies is the key and difficult point of relevant disciplines at present.Runoff forecasting and reservoir dispatching are typical optimization problems.With the development of modern computer technology,swarm intelligence algorithm has shown efficient optimization performance in the field of optimization.Applying swarm intelligence algorithm to runoff forecasting and reservoir dispatching is a hot research topic in related fields.Based on the idea of level,this paper improves the optimization performance of firefly algorithm through multi-strategy integration,and applies the improved algorithm to runoff forecasting and reservoir dispatching respectively.The main work of this paper is as follows:(1)A firefly algorithm based on level-based learning and variable step size is proposed.The level-based learning model is used to replace the full-attractive model.The firefly particles are sorted by brightness and divided into several layers.Two firefly particles are randomly selected from the better layer for each firefly particle to learn.Using the variable step size strategy,the compromise length decays with the iteration of the algorithm.The level-based learning strategy enables the firefly particles at different levels to have different development and exploration abilities,thus effectively balancing the development and exploration abilities of the algorithm.The variable step size strategy can meet the needs of the algorithm in different stages of evolution.(2)A firefly algorithm based on single-dimensional self-learning is proposed.Firefly particles are divided into two parts.Firefly particles with good adaptive value randomly choose a firefly particle other than itself from the population,and randomly choose a dimension to produce three candidate solutions,and then greedily choose the best solution among the three candidate solutions and themselves.Two firefly particles with poor adaptive value are selected to learn at the same time.The use of both learning approaches greatly avoids the pitfalls of using a single approach.Finally,a runoff forecasting model based on firefly algorithm is constructed,and the improved algorithm is applied to runoff forecasting to improve the accuracy of runoff forecasting.(3)A firefly algorithm with multi-strategy division of labor is proposed.According to their performance,firefly particles are divided into leaders,developers and followers,and different strategies are assigned for firefly particles according to the different division of labor.The leader carries on the greedy Cauchy mutation;The developer randomly chooses two leaders to use elite neighborhood search strategy for local development.The follower randomly selects two excellent firefly particles for global exploration.The multi-strategy division of labor enables the leader to break out of local optimal,and at the same time can preserve the excellent information of the population.The developer and the follower are respectively responsible for the development and exploration of algorithms.Finally,a reservoir dispatching model based on firefly algorithm is constructed to improve the reservoir generating capacity.