Analysis Of Power Load Characteristics And Short-term Power Load Forecasting

The establishment and construction of smart grid has become an important development and research object of the State Grid at present,because the efficien cy of grid operation is improved,saving power,environmental protection,optimizing power system assets,and improving the quality of power supply services are all due to the establishment of smart grids.It is getting better and better,but at the same t ime the power system has put forward higher requirements for all aspects.Short-term load forecasting of power systems is an important basis for the safe operation and economic development of power systems.Therefore,the power sector has put forward higher requirements for more accurate short-term power load forecasting.Because the power load is affected by many uncertain factors,short-term power load forecasting is performed by using a clustering method to select similar days to improve prediction accuracy and speed.For the influence of various data of climate load on the electric load,numerical values such as temperature,humidity,weather,type of week,season,etc.,which cannot be digitized,are established by establishing a mapping database,so that different influencing factors are numerically comparab le..In terms of prediction,support vector machine is the mainstream prediction method in machine learning.It has many unique advantages in dealing with small sample data,high dimensional recognition and nonlinear problems.The specific research conten ts are as follows:(1)Firstly,it analyzes the characteristics of electric load,comprehensively and systematically introduces the current research status of short-term electric load,analyzes the characteristics of electric load itself,and then analyzes the influencing factors of electric load to establish the factors affecting the change of electric load.Study and analyze it to determine the establishment of the model.(2)Secondly,for the influencing factors analyzed,determine the influencing factors to be added to the model,and select the sample data of the short-term electric load forecast that meets the conditions.A feature quantity mapping database is established,and similar data with similar characteristics to the prediction date are selected according to the clustering method,and the data is filtered to form predicted sample data.(3)Thirdly,based on the sample data after screening,a regression prediction model is established,and the AFSA-SVM model is used for short-term electric load forecasting.By analyzing the examples,more accurate prediction results are obtained.And use the same initial sample data to model and predict the analysis using other methods,and compare it with the final result.The prediction effect of this method was found to be better.Through the final experimental analysis and comparison,it is found that the method of selecting the similarity day to select the sample data by clustering and then using the ASFA-SVM model for prediction and prediction can effectively introduce various factors affecting the power load.It can improve the prediction accuracy of short-term power load more accurately and shorten the prediction time.