Research On Techniques Of The Medium And Long Term Power Load Forecasting For Power Network Planning

Grid planning is an important basis for the construction of power grids and development on power grids.Scientific and effective planning is of great significance to guide the construction of power grids.The medium an long term load forecasting is the basic link of power grid planning,and its accuracy will directly affect the safe and reliable operation on the power grid planning.The accuracy is of great significance to economic and social benefits on all aspects of the power grid planning,construction and operation.In response to the small sample feature of the medium and long term load forecasting and the limitation problem that the ordinary least squares(OLS)method can not effectively deal with problems of the high degree multiple correlation in the independent variable system.In this paper,the basic principle and its contents of the partial least squares regression(PLS)model are introduced,and the steps of the algorithm are deduced in detail.In order to solve the problem that the partial least squares regression method can not be used to optimize the input variables,the sparse partial least squares regression(SPLS)model is introduced,which mainly adds the variable selection process on the basis of PLS.The medium and long term load forecasting model based on sparse partial least squares regression is established by introducing the steps of sparse partial least squares regression algorithm and the basic content of its parameter selection.By comparing the SPLS with OLS,PLS models,it is proved that the sparse partial least squares regression model shows more superiority and accuracy on processing the medium and long term load forecasting.Considering the multiple factor comprehensive influence,there are horizontal and vertical limitations when the traditional prediction algorithm model is used to the medium and long term load forecasting.The main reason is that failure to scientifically and effectively optimize the input of the relevant variables lead to lower prediction accuracy,which is difficult to strike a balance between considering more influencing factors and reducing the prediction model errors.In addition,the input of the complex and diverse variables,redundant information is not preprocessed,which leads to the low training performance of the subsequent algorithm model and the slow calculation speed.To this end,the rough set method(RS)is introduced to be applied to the medium and long term load forecasting of multi-index model.In this paper,the basic idea of rough set is given and the rough set attribute reduction algorithm based on attribute importance is described in detail.The rough set is combined with the traditional algorithm model to improve its performance.The rough set attribute reduction algorithm is used to reduce the input variables for the original data processing,to eliminate the relatively unnecessary attributes of the model,to simplify the input complexity of the model,and to scientifically and effectively select influencing factors such as high degree of influence and high correlation from the many factors that affect the power load.On the basis of the rough set preprocessing procedure,the medium and long term load forecasting model based on rough set is established by respectively using partial least squares regression and support vector machine method after obtaining the optimal attribute set samples which are suitable for the prediction object.Then regression model based on the rough set(RS-PLS)and the support vector machine model based on rough set(RS-SVM)are respectively established for the medium and long term load forecasting.The example is given to show that the introduction of rough set method provides the theoretical basis for the choosing influencing factors of input variables,which avoids the defects of human choice by subjective experience.On the other hand,it simplifies the complexity of model input,accelerates the training speed and reduces the training steps and time.The accuracy and stability of the medium and long term load forecasting model based on rough set are validated,and it is more objective and scientific when the rough set method is applied to the medium and long term load forecasting considering multiple factor comprehensive influence.