Enterprise Daily Electricity Consumption Forecasting Model For Smart Grid Based On Big Data

With the continuous development and integration of big data technology and power information technology,more and more machine learning algorithms have also been applied to the smart grid field,and power demand forecasting has become a hot topic.The traditional power system is dominated by the top-down operation management mode,which often results in a certain waste of resources because the user’s power demand does not reach the usage.However,in today’s smart grid system with integrated,high-speed two-way communication network,hardware and objective conditions such as weather and temperature can be collected through advanced equipment technology,control methods,sensing and measurement techniques.It is converted to continuous or discrete digital features.Then use this data as the training set,and use the prediction algorithm to debug the appropriate model to predict the boundary value of the user’s power demand range,thus strengthening the power management on the power demand side,so that the power resource can be used more efficiently.At present,the main domestic electricity consumption forecasting direction is often based on the overall electricity demand of a certain region as the forecasting target to meet the overall electricity consumption planning,and this paper focuses on the specific electricity users of the enterprise to predict the electricity consumption.It is hoped that an accurate model of abnormal power consumption behavior can be constructed by means of accurate electricity demand forecasting method,so as to further deepen the forecast of the electricity demand of enterprises.This can not only solve the problem of power demand of enterprises,but also reduce the waste of power resources,so that the distribution of power can be balanced.This paper takes the daily electricity consumption of 1454 companies in Yangzhong City as the sample data set.In the first three chapters,it mainly analyzes the characteristics of the data set,normalizes the data,and proposes innovative data through data visualization.The enterprise level method is divided according to the power consumption level,and the feature mining is carried out on the inside of the data set to construct the feature engineering.In the fourth chapter,the power prediction model is built based on linearregression algorithm,and the simulated annealing algorithm is used for feature screening.In the fifth chapter,the XGBoost algorithm is used to build the power forecasting model,and the filtering method in the XGBoost algorithm is used for feature screening.The final prediction effect of the model is good,and it can be done for more than 97% of large,medium,and medium-sized enterprises.Offset error is less than 10%,74% of medium-and medium-sized enterprises have offsets less than 5%.For small enterprises,85% of small-scale enterprises can also achieve offsets of less than 10%,which basically meets actual forecasting requirements.