Research On Short-term Load Forecasting Based On Improved Cat Swarm Optimization Neural Network

Power load forecasting is an important part of power system security and economic development planning.It is very important for power system planning and operation,and it is also a prerequisite for dispatching,designing and researching.In the context of the country’s vigorous promotion of wind power,photovoltaic and other new energy power generation,short-term load forecasting is very necessary in view of the unstable power of new energy power generation such as wind power and photovoltaic power generation.BP neural network is has very strong processing ability in terms of nonlinear data,and it is chosen as the basic model of short-term load forecasting here.Because the BP neural network has the shortcomings of long training time and difficulty in selecting suitable initial values,this paper combines the cat group algorithm with it and conducts research.Test analysis shows that compared with ordinary single algorithms,the results are obvious Optimization.The unique search mode and tracking mode cooperation effect of the CSO ensures the algorithm’s global search ability.The convergence speed of the algorithm is faster in the initial stage of the iteration.As the number of iterations increases,if the distribution of the CSO is not adjusted,some cats will still execute the global search wastes resources,and the accuracy of the local and global optimal solutions improves slowly.Based on the introduction of the quantum mechanics model and the CSO,the method of selecting the grouping rate and the individual in the search mode are introduced.The improvement of the copy formula optimizes the allocation mode and search mode,thereby improving the search efficiency and accuracy of the algorithm.This paper uses three different forecasting models to study the load forecast in a certain area of our country,and conduct simulation experiments on matlab.The simulation data demonstrates that the improved BP neural network model optimized by CSO is superior to the ordinary BP neural network model and the BP neural network model optimized by the standard CSO in terms of prediction accuracy,which proves the feasibility and scientific nature of the method proposed in this paper.