Research On Day-Ahead Load Forecasting Model For Electricity Sellers

With the publication of the "Several Opinions on Further Deepening the Reform of the Power System" by the Chinese government in 2015,the reform of the domestic electricity market has also intensified,and a large number of market participants such as power sales companies have participated.In this context,the provincial governments have issued a series of bias assessment systems to properly guide the direction of power market reform,and stipulate that the corresponding penalties must be imposed in the deviation of electricity.In general,the power sales company is also obliged to bear the risk of deviation assessment caused by fluctuation in electricity price and power load forecasting while participators earn high returns in the power market transaction earn high returns,aiming to maintain the stable development of the market.Although domestic scholars have made great achievements in the research of load forecasting at this stage,with the advancement of the power market reform process,a series of new research problems have emerged,such as how the electricity sellers can predict the load more accurately in the future.In order to better deal with the risk of deviation assessment and there are also the influence of location factors on the accuracy of the forecasting when the companies sell electricity in different regions.Problems like this need to be learned to explore and solve.This paper mainly explores a series of bias assessment problems faced by electricity sellers in different regions under market reforms,which leads to the establishment of a related model to forecast load.Facing the above demand,multiple algorithms are applied to realize the multiple parallel forecasting of load,and then establish the data fusion model to fuse the multiple forecasting results into one final result.The pre-prediction work is based on the clustering of historical load data,and establishes a pattern recognition model based on SVC.Then the load data of different regions are randomly divided into different parts,and the load forecasting is carried out with the basic model respectively.The load forecasting result is used as the input of the pattern recognition model,and the corresponding prediction mode is obtained.Because the load of different regions has different trend fluctuations,but there are intrinsic links,the results of load forecasting in each region are directly added,which does not reflect the relationship between these influencing factors.Therefore,this paper proposes day-head load forecasting model based on pattern recognition for adaptive time section fusion.In this model,the mutual feedback of the fusion model classification model and the data fusion model is also added,which further improves the accuracy of day-head load forecasting.