A Short-term Load Forecasting Method Based On Spark And Holt-Winters Model

Short term load forecasting is an important guarantee for economic operation and dispatch of power system.With the development of smart grid and sensor technology,the power grid data is expanding violently.The short-term load forecasting under the condition of large power grid data requires a high prediction accuracy and a fast calculation speed.The traditional forecasting methods are unable to meet the needs.With the emergence of big data processing technology represented by Hadoop and Spark platform,the way to deal with big data has entered a new stage.In this paper,the key problem is to improve the prediction accuracy and improve the short-term load forecasting speed.Firstly,according to the prediction and experimental short-term load,using the load data of NREL labs announced,generated about 25 million group load test data set;The characteristics of commercial electricity load and household electricity load are analyzed,and two types of loads have periodic and seasonal characteristics.Then according to the features of load data,build a short-term load model based on the multiplication Holt-Winters method,and apply the L-BFGS algorithm in unconstrained optimization theory to the model parameters optimization to reduce the computational complexity and to ensure the accuracy of load forecasting.Then,the short-term load forecasting algorithm is implemented in parallel with the Spark cluster to improve the calculation efficiency and realize the prediction of mass load data.A cluster construction scheme with 2 Master nodes and 28 Slave nodes is proposed,and the configuration of the cluster is optimized so that Spark computing performance is better.Finally,The HDFS load data storage in this paper is more efficient than the traditional file storage mode;The L-BFGS optimization method has good computational efficiency and optimization accuracy;Compared with the traditional load forecasting algorithm,the Holt-Winters algorithm used in this paper has higher prediction accuracy;The short-term load method after Spark parallel implementation can cope with the prediction requirements of mass load data.In the computational cluster of 25 computing nodes,the proposed method can achieve 2 million scale load forecasting in 1.5 minutes,and realize 20 million scale load forecasting in 13 minutes.The method can reduce the cost of power system,save more time,and provide the guarantee for the dispatch and control of power system.In this paper,the short-term load forecasting method based on Spark and Holt-Winters is a feasible solution for mass short-term load forecasting.