A Research Based On Hybrid Methods GOA And LSSVM For Electric Load Forecasting

With the implementation of the new round reform in power system,substantial progress have been made in the past three years.In the new era,high quality development has become the fundamental requirement for the reform and development of the power industry.Ensuring a reliable power supply and providing a better service at the same time are significant for safeguarding and improving people's life quality.An accurate and reasonable prediction of power load can help electricity companies to arrange power generation,and adjust the transmission plans reasonably as well.It is conducive to rationally arrange power grid maintenance and infrastructure upgrade for power grid enterprises.However,the power load is affected by many factors,such as temperature,season and economic situation,which will affect the prediction results,and increase the difficulty of power load prediction.This thesis is based on the related literature,and study of the current situation of power load predict.We have studied the related prediction methods,especially the intelligent optimization algorithms and least squares support vector machine algorithm.The specific research content and achievements of this thesis are as follow:(1)An improved grasshoppers optimization algorithm.The traditional grasshopper optimization algorithm sometimes will fall into the local optimal solution.To overcome this problem,we combine the differential evolution and grasshopper optimization algorithms,a new method IGOA is proposed,and apply it on parameters optimization.(2)Design the combined forecasting model(WS-IGOA-LSSVM)base on the wavelet threshold denoising in the optimized least square support vector machines(LSSVM).For the randomness and dynamics of the short-term power load,apply the wavelet shrinkage method in electric load data for noise eliminate,then in the least squares support vector machine parameter and kernel function parameters to solve the problem,using LSSVM model to optimize the parameters of grasshopper improved optimization algorithm.We finished an experiment with the history load data of New South Wales and Queensland two areas in Australia,the experimental results show that the proposed model have a nice prediction performance in short-term electric load.(3)Proposed a method Mixed-PLSSVM which is based on the combination of the sparse kernel least squares support vector machine and Mixed-kernel function.Because the mid-long term load has less historical data,and it is also influenced by many factors,such as national economy,population and so on,select the influence factors with strong correlation will improve the accuracy of prediction.The model firstly analyze the factors for total power consumption data and related influencing factors,select the correlation factors to form a data set,then combines the RBF kernel functions and the variant chi-square kernel functions,and adopt an improved optimization algorithm to optimize the relevant parameters of the sparse least squares support vector machine in mid-long term load forecasting.The experimental results show that the proposed method has good prediction performance in mid-long term load forecasting.