Research And Implementation Of Power Grid Operation Data Analysis System Based On Spark Streaming

The rapid development of information technology has enhanced the depth of its application in various industries.As an important link of energy conversion and transmission,the power grid has become an important link for the industrialization and economic development of the whole country.After the epidemic entered a new stage,the economy entered a period of rapid recovery,the number of various terminals connected to the power grid increased,and the traditional power grid management mode could no longer adapt to the current scenario.In this scenario,the new technology of smart power grid is adopted to integrate and manage the current big data,Internet of things,intelligent control and other technologies,so as to achieve efficient and safe management and control of the entire power grid and improve the informatization degree and ability of the entire power grid.The main work of this paper includes the following aspects:(1)Power grid operation data analysis system In order to solve the problems of slow network speed and low transmission efficiency in the process of data information collection in the past,the emerging Internet of Things technology,mainly low-power wide area network technology,is adopted to improve the speed and efficiency.On the basis of this technology,information sensing equipment selection,intelligent control process design and equipment integration and other operations are carried out to strengthen the monitoring of the transformer body and environmental data,and more accurately master the temperature appreciation and load coefficient of the distribution transformer,which improves the economy and applicability,but also provides a basis for later equipment maintenance and load state analysis.Realize the perception of power grid operation data information and remote intelligent control of equipment,so as to make the whole power grid.(2)A GRU algorithm based on attention mechanism is proposed.The processing of the whole algorithm fully takes into account the periodicity,tendency and relevance of the data.The whole processing architecture is divided into five layers,which can complete the pre-processing of historical charge data information.The CNN algorithm is used to complete the extraction of input features,dimensionality reduction analysis and other operations,and the attention mechanism is used to enhance the influence of important information.In this way,the information analysis and prediction of power grid load data can be completed.Combined with the real load data information of the whole city of Yangzhou,the simulation results show that the prediction error of the GRU algorithm combined with attention mechanism is significantly reduced compared with the traditional neural network algorithm,and the calculation accuracy is significantly increased.(3)Big data analysis:Various business data information will be generated at different stages of power grid operation.In order to give full play to the value of these data information,the system adopts data mining algorithm for big data analysis.Spark Streaming,a distributed real-time computing framework,is applied to the condition monitoring system.The power grid load is predicted by convolutional neural network,and the trained model is saved as PMML model file,and then the trained model is loaded by Spark Streaming for real-time load prediction.The power grid operation data analysis system designed based on information technologies such as the Internet of Things,big data and cloud computing can not only ensure the accuracy of system data,but also provide digital support for the entire power grid operation decision by using big data analysis and intelligent processing technology.