An IPSO-BP Neural Network For Medium-term Power Load Forecasting Based On Wavelet Analysis

Electric industry is the foundation of a country’s industrial development,which is related to national security,social stability,and people’s livelihood.Power load forecasting is the basis of safe operation and reasonable planning of power systems.Accurate power load forecasting is conducive to reduction of power generation costs and improvement of operational efficiency.With the rapid development of the power industry,the traditional forecasting methods with defects in the prediction accuracy and time can no longer fully meet the forecast requirements of the actual power system.The upgrade and progress of the prediction technology have been put on the agenda.In this paper,firstly,the influence factors of power load are studied,including the impact of time series disturbances,weather changes,economic development,and industrial structure on power load.A complete variable quantification standard is specified for the prediction model.Based on BP neural network,aiming at the defects of the existing forecasting methods in forecasting accuracy and time,we propose an improved scheme based on the following methods and construct an improved power load forecasting model.First,the wavelet transform is introduced to decompose the load sequence.The neural network model for each fundamental wave and harmonic component is built in parallel.The prediction results are calculated based on the wavelet superposition principle to improve the training speed and prediction accuracy.Second,the PSO algorithm is improved.The IPSO algorithm is used to optimize the initial value of weights and thresholds of each neuron in the neural network,so as to reduce the probability of falling into the local optimum and improve the training speed.Third,based on the node correlation,the multi-time node prediction model is established to reduce the sample set size of neural network input and the influence of time series on the prediction model.In this paper,the power load data of a certain region in North China are used to construct the medium-term power load forecasting model.The improved IPSO-BP model and the traditional BP model are constructed respectively,and the comparison test is conducted between the prediction results and the actual value.The experimental results show that the speed and accuracy of the IPSO-BP method are significantly improved compared with the traditional method,and the feasibility and effectiveness of the improved method have been fully verified.