The Research On Wind Power Prediction Based On The IPSO-BP Neural Network Model

The wind's fluctuation and intermittent could have great influence on the power quality and stable operation of the power system.Wind power prediction is one of the key technologies to solve this problem,therefore the research of wind power prediction method is of great significance.The BP neural network can be one of the ways to predict the wind power,it has its own strength,such as strong nonlinear mapping ability and fault tolerance ability,etc.But the weakness of the BP network is also prominent,such as slow convergence speed,easily trapped in local minimum,etc.So an improved particle swarm optimization algorithm(IPSO for short)is put forward to optimize the BP network in this paper,then the optimized network is used to forecast the short-term wind power.The main research content of this article are as follows:(1)A linear decreasing inertia weight is used in the standard particle swarm optimization(PSO for short).But in the early stage of the iterations,its global search ability is not strong,and in the later stage,its local search ability is general.To improve both the global search ability and the search speed and precision of particle swarm optimization algorithm,the corresponding improvement methods are present to improve the inertia weight,accelerating factor and the particle velocity boundary respectively.The Inertia weight uses nonlinear decreasing method,in the early iterations with larger inertia weight,such adjustment can improve the global search ability and in the late iterations with a smaller inertia weight,such change can improve the local search ability and search accuracy,speed up the convergence speed.For two acceleration factor,the first one has a linear decrease,the another has a linear increase.the speed of the particle boundary uses a linear decreasing way.The two improvements are beneficial to the global search of the early iterations and local search of the later iterations.Then four functions are used to test the convergence of the IPSO algorithm.The test result shows that the IPSO algorithm is is superior to the PSO algorithm on the convergence speed and accuracy.(2)Aimed at the defects of BP neural network,the IPSO algorithm is applied to the BP neural network to optimization the network's weight values and threshold values,then the optimized network is used to establish prediction model to predict the short-term wind power.the training sample and test sample are respectively used to train and test the three networks.the experimental results show that compared with the BP network and optimized BP network by PSO algorithm,the BP network optimized by IPSO algorithm has faster convergence speed and better prediction effect on the short-term wind power.The experimental results testify the effectiveness of the prediction scheme in this paper.