Research On Daily Electricity Consumption Forecast Of Enterprises Based On Differences In Different Industries And Neural Network Algorithm

In the context of changes in global coal supply and the national strategy of “peak carbon,carbon neutrality”,the surge in domestic electricity demand has led to a tight supply of electricity.Accurate forecast of daily electricity consumption of enterprises can provide decision-making basis for the market-oriented floating mechanism of electricity price and demand response scheme,thus guiding enterprise users to reasonably adjust electricity demand and relieve electricity shortage.Therefore,high-precision prediction of daily electricity consumption of enterprises has important practical significance and application value,no matter for the operation and development of power supply companies,power plants or electricity sales companies.Based on the full analysis of domestic and foreign research on electricity consumption forecasting,this paper proposes the idea of adopting differentiated modeling for enterprises in different industries according to the cluster analysis of the monthly electricity consumption of enterprises and the related factors of daily electricity consumption,so as to improve the forecasting accuracy.For this purpose,the traditional neural network and deep learning neural network are used to realize the forecast of daily electricity consumption of enterprises.The main research contents of this paper are as follows:(1)Take the electricity consumption data collected by all enterprises in the urban area under the jurisdiction of a power supply company as the original data,and use the K-means algorithm to perform cluster analysis according to the monthly electricity consumption,and divide the enterprise users into different clusters according to the electricity consumption level and electricity consumption law.And analyzed the characteristic attributes of each cluster.Based on the conclusions of the cluster analysis,the classification of different industries,factors affecting the daily electricity consumption of enterprises,and the determination of the input characteristics of different industries are discussed and analyzed,and the data basis and model structure settings are provided for the differential establishment of the daily electricity consumption forecast models of enterprises in different industries.in accordance with.(2)Constructed the neural network(Back Propagation,BP)of error back propagation to predict the daily electricity consumption model of enterprises in the whole industry,and the radial basis neural network(Radial Basis Function,RBF)to predict the first category(with production planning nature)enterprises The daily electricity consumption model uses the powerful feature extraction ability of convolutional neural network and long short-term memory neural network to describe the long-span better in view of the low accuracy rate of forecasting the daily electricity consumption of the first-class enterprises by BP and RBF networks.The ability to change time series data realizes the initial model(Convolutional Neural Networks＿Long-short Term Memory Networks,CNN＿LSTM)combining convolutional neural network and long short-term memory neural network,which improves the prediction accuracy of daily electricity consumption of enterprises with production planning nature.(3)In order to further improve the prediction performance of the CNN＿LSTM network,the structure optimization and parameter debugging of the CNN and LSTM architectures of the initial model were carried out,and the simulation results of the examples under the same and different operating modes were compared,and it was verified that the prediction of the CNN＿LSTM model has the nature of production planning The daily electricity consumption of enterprises has higher precision and stronger generalization ability.This paper integrates the prediction performance of the daily electricity consumption models of enterprises in different industries,realizes the application of the daily electricity consumption forecast of a power supply company,and provides data support for the real-time bidding and demand response of electricity transactions.And in view of the low accuracy of the power supply company’s traditional experience in asking the enterprise’s electricity consumption,this paper uses the neural network algorithm to achieve a prediction accuracy of 90%,which improves the accuracy of a power supply company’s prediction of the enterprise’s electricity consumption.