Analysis And Forecasting Of Load Curve Of Distribution Transformer

With the continuous improvement of smart grid construction,a large amount of power data has gradually accumulated in the power system.This paper takes the load data of distribution transformers as the research object,uses data mining and other related technologies to study the change law of load curve of distribution transformer,summarizes the daily load patterns of distribution transformers,and analyzes the similarity of power consumption behavior between distribution transformers.These studies are helpful for the research of power big data such as demand response,load forecasting,and electricity price design.At the same time,in order to grasp the future trend of the load curve of distribution transformer,the paper use deep learning and other related technologies to study short-term load forecasting methods of distribution transformer.Accurate load forecasting provides important support for smart grid and efficient energy management.The main contents of this article are as follows:First,this paper proposes an adaptive load curve clustering method based on singular value decomposition and combines individual to overall analysis method to achieve large-scale distribution transformer daily load curve clustering,and extract daily load patterns of distribution transformers.In order to further explore the variation law of load curve,this paper proposes a classification method for distribution transformers based on similar electricity consumption behavior.This method divides distribution transformers with similar daily load patterns and similar daily load pattern changes into one category.The classification conclusion obtained using the proposed method provides support for cluster division in short-term load cluster forecast of distribution transformers.Secondly,in order to reduce the influence of data distribution and network structure of the prediction model on the accuracy of short-term load forecasting,this paper proposes a hybrid prediction model that includes STL,LSTMs,and XGBoost regression.Through multiple sets of comparative experiments,it was verified that adding sequence decomposition and ensemble learning could improve the performance of the short-term load forecasting model.Due to the large number of distribution transformers,in order to avoid training the prediction model for each distribution transformer,this paper proposes a short-term load cluster prediction method of distribution transformers based on similar power usage behavior.Through experimental comparison,the proposed method can guarantee the prediction accuracy and achieve the cluster prediction of multiple distribution transformers.